Antecedents of Gender Harassment: An Analysis of Person and Situation Factors
- First Online:
- Cite this article as:
- Hitlan, R.T., Pryor, J.B., Hesson-McInnis, M.S. et al. Sex Roles (2009) 61: 794. doi:10.1007/s11199-009-9689-2
- 230 Views
Two laboratory studies examined the impact of person and situation factors in the prediction of gender harassment. Male undergraduates from a mid-sized Midwestern university in the U.S. were recruited based on an assessment of sexist attitudes. It was predicted that, across two studies, characteristics of one’s personality (sexist attitudes) and situational characteristics (sexual priming and masculine identity threat) would produce unique and interactive effects in the prediction of gender harassment, defined behaviorally as the number of sexist questions asked of women during a mock job interview, and cognitively as negative evaluations of the female interviewee. Across both studies, results support the predictions that both person and situation factors are important to understanding gender harassment.
KeywordsGender harassmentPerson x situation analysisSexual harassmentSocial identity
From the earliest to the most contemporary surveys, sexual harassment has been found to be a predominantly male behavior directed at predominantly female targets. The most common form reported is gender harassment (Bastian et al. 1996; Fitzgerald et al. 1995; Gutek 1985; USMSPB 1995). Unlike other forms of sexual harassment, gender harassment is not aimed at gaining sexual access to women; rather, it is an expression of ridicule—it represents a put down (Gelfand et al. 1995; Pryor and Whalen 1997). One of the most common examples of gender harassment is sexist verbal behavior. For example, Bastian et al. (1996) reported that "[unwanted] sexual teasing, jokes, remarks or questions" were experienced by 44% of military women across all branches of the U.S. armed forces. Gender harassment is also a frequent experience among women in nontraditional, especially blue-collar jobs (DeCoster et al. 1999; Ragins and Scandura 1995).
The United States Merit Systems Protection Board (USMSPB 1995) reported that gender harassment was the only form of sexually harassing behavior that showed a steady increase across the administrations of a survey of federal employees that began in 1980. When other forms of sexually harassing behavior occur, such as unwanted sexual attention or sexual coercion, they tend to be accompanied by gender harassment. In contrast, gender harassment is likely to occur in isolation from other forms of sexually harassing behaviors (Fitzgerald et al. 1999). Perhaps because it is more commonplace, gender harassment is often viewed as less severe than other forms of sexually harassing behavior. Nevertheless, research has indicated that gender harassment is related to both negative personal and professional consequences for women (Bowling and Beehr 2006; Gutek 1985; USMSPB 1995; Parker and Griffin 2002), including diminished psychological well-being and a variety of psychological stress symptoms (Carr et al. 2000; O’Connell and Korabik 2000; Piotrkowski 1998; Woodzicka and LaFrance 2005).
Why do men perpetrate gender harassment upon women? Pryor and his colleagues have theorized that sexual harassment has both person and situational antecedents. In other words, some people may be predisposed to sexually harass and some social situations may be conducive to sexual harassment (Pryor and Fitzgerald 2003; Pryor et al. 1995). The goal of the current research was to examine some of the personal characteristics and social situational factors that facilitate gender harassment. We asked undergraduate men to serve as interviewers within a common organizational context—a job interview. Across two studies, we examined the importance of various situation and person factors to the manifestation of gender harassment. Specifically, Study 1 examined how sexist priming (a situation factor) and sexist attitudes via the modern sexism scale (a person factor) contribute to gender harassment. Study 2 further examined the additional situational factor of masculine identity threat along with sexist attitudes as antecedents to gender harassment. It was predicted that more sexist men would be more likely to behave in a harassing way toward a female interviewee. We also predicted that priming sexist views of women and threatening men’s masculine identities would increase the likelihood of gender harassment. Finally, we predicted that sexist attitudes would interact with situational forces in predicting gender harassment. Below we outline the rationale for our operationalizations of both the person and situational factors in these studies.
Characteristics of Men who Commit Gender Harassment
We measured individual differences in sexism as a possible antecedent of gender harassment using the Modern Sexism Scale (MS; Swim et al. 1995). According to Swim et al., the MS reflects subtle and covert negative attitudes about women and avoids questions about more obvious traditional gender roles and negative stereotypes of women found in measures of Old Fashion Sexism (OFS). In this way, the MS is less subject to political correctness biases that skew responses to other measures of sexism. Similar to work on modern forms of racism (McConahay 1986; Sears 1988), the MS scale includes items assessing denial of continued discrimination, antagonism about women’s demands in contemporary society, and lack of support for policies designed to support the rights of women. Higher MS scores are correlated with greater use of sexist language, overestimating the percentage of women in male-dominated occupations, and a decreased likelihood of viewing discrimination as resulting from sexism (Swim and Cohen 1997; Swim et al. 2004). In both of the studies described below, we predicted that men scoring higher in MS would be more likely to engage in sexist behavior and to evaluate females more negatively.
Situational Factors and Gender Harassment
Research on sexual harassment has also indicated that several situational factors may contribute to gender harassment (for a review see Pryor and Fitzgerald 2003). One factor that seems to cross-cut many social situations is the degree to which environmental cues prime sexist stereotypes (Banaji and Hardin 1996; Blair and Banaji 1996). For example, Rudman and Borgida (1995) exposed men to television commercials in which women were portrayed as sex-objects. During a subsequent mock interview task, primed men were more likely to encode information about and behave toward a woman in a sexualized way (e.g., pay more attention to her appearance and her style of dress than what she said, ask more sexist questions, sit closer to her, and spend more time gazing at her body (see also McKenzie-Mohr and Zanna 1990). Thus, we expected that gender stereotypic priming would result in greater levels of gender harassment and more negative evaluations of women.
Another situational factor that has been suggested as a potential contributor to gender harassment in traditionally male workplaces is masculine identity threat. For example, Gruber and his colleagues (Gruber and Smith 1995; Gruber and Bjorn 1982) suggested that men in blue collar jobs may feel threatened by women who have similar jobs and earn similar wages. Gruber maintains that the proneness of men in traditionally masculine jobs to harass women may represent an attempt to put women in their place or to regain the upper hand. Similarly, Martin (1990) suggested that male police officers often fear that female police officers who perform at a level equal to or better than their own undermine the sense of masculine identity associated with police work. Consistent with this reasoning, Martin found that approximately 75% of female police officers had experienced sexual harassment on the job.
Harassment experiences of women in traditionally masculine jobs seem to reflect an “us versus them” perspective on the part of men. Extensive research on Social Identity Theory (SIT; Tajfel and Turner 1986) indicates that those categorized as out-group members are rewarded less, perceived to have fewer positive attributes, and perceived to be more homogeneous than in-group members, and this appears to be especially true when an out-group is perceived as a threat to one’s in-group (Brewer and Brown 1998; Gaertner et al. 1989; Hamilton and Sherman 1994; Linville et al. 1989). In-group/out-group categorizations may automatically activate evaluative structures that can influence perceptions of group members (Dovidio and Gaertner 1993). While the preference for in-group members does not necessarily lead to hostility toward out-group members, the need to justify superiority of the in-group over the out-group can lead to disdain and hostile behavior (Brewer 1999). Such disdain may, in turn, lead to devaluing the performance of out-groups, especially when they out-perform the in-group (Schmader and Major 1999). To the extent that men view other men as members of an in-group and women as a distinct out-group, they may be particularly likely to marginalize women, thereby enhancing their own sense of gender identification (cf. Maass et al. 2003). In fact, Maass and her colleagues suggest that threatened social identity may be one of the primary motivating factors involved in gender harassment.
The current research investigates how masculine identity threat might increase the likelihood of gender harassment. Consistent with SIT, such a threat should result in women being viewed more as members of a distinct out-group and subject to behaviors and attitudes commensurate with such a perception. We predicted that when men’s gender identities were threatened by women who out-performed them on a male-gendered task, they would be more likely to engage in gender harassment and exhibit negative evaluations of women.
Person X Situation Analysis
As noted above, some of the antecedents of gender harassment may take the form of individual proclivities to perform certain types of behavior, whereas other antecedents may involve social situations that make certain behaviors more likely to occur (Gruber 1998; O’Connell and Korabik 2000; Welsh 1999). This lens through which the causes of sexual harassment may be viewed has been called the Person X Situation Model (Pryor et al. 1993, 1995; Pryor and Whalen 1997). Pryor and his colleagues conceived of sexual harassment as a behavior that some men do some of the time. More specifically, sexual harassment is most likely to occur when men with a proclivity to sexually harass are placed in a situation amenable to such behavior. Original evidence for the Person X Situation Model came from Pryor et al. (1993) who found that sexually harassing behavior (i.e., unwanted sexual attention) occurred when men who were high in the Likelihood to Sexual Harass (LSH) were exposed to an authority figure who displayed such behavior himself. Thus, the interaction between person and situation factors was crucial to produce sexual harassment.
While not all research consistently shows such an interaction effect (see Rudman and Borgida 1995), several studies suggest the importance of both personal and situation factors. For example, Dekker and Barling (1998) conducted a survey of Canadian university faculty and staff using men's retrospective self-reports of having engaged in gender harassment. Results indicated that men with certain personality characteristics (e.g., weak perspective taking skills, adversarial sexual beliefs, inappropriate sexual harassment beliefs) were more likely to report engaging in gender harassment than others without such characteristics. Of particular import, these men were most likely to admit such behaviors when company sanctions for such behavior were perceived as weak. Similarly, Maass et al. (2003) found that men’s gender identification and Social Dominance Orientation (SDO) interacted with situational factors, such as perceived threat (operationalized as whether a woman described herself as being traditional or non-traditional in her gender-role beliefs) such that men who more strongly identified as male were particularly likely to harass when the woman described herself as being non-traditional.
To wit, we view gender harassment as a behavior produced by a complex interplay between one’s personality and the social situation. Consistent with this line of reasoning, we expect that personality may function to constrain behavior (cf. Colbert et al. 2004). Following this constraint hypothesis, we argue that individuals with certain personality characteristics (e.g., low sexism, low LHS) are more constrained in their behavior and expressed attitudes across different types of social situations. For example, given the relations between sexist attitudes and other gender-related attitudes, we would expect that low sexist men would be more empathetic to how sexist behaviors are viewed by women and less likely to behave in a sexist way across situations. However, at high levels of sexism, we expect personality to exert less constraint on behavior. That is, the behavior of men with more sexist attitudes should be more susceptible to their immediate social-situational influences.
The Present Research
Compared to other forms of sexual harassment, gender harassment is inherently more hostile in nature and is used as a reaction to women who are viewed as distinct out-group members. Such hostility seems especially likely among highly sexist men. Across two studies, men's scores on the Modern Sexism Scale were examined as possible predictors of gender harassment. As we have reviewed, situational factors that prime sexist stereotypes or threaten men’s masculine identity could also increase such hostility. Thus, Study 1 examined whether the presence of sexist priming was related to acts of gender harassment. Similarly, Study 2 examined whether men whose sense of masculine identity was threatened by a woman who out-performed them might be more likely to engage in gender harassment. Both studies also examined whether personality (sexist attitudes) moderated the effects of situational factors (priming, identity threat) on gender harassing behavior and attitudes. Gender harassment was defined behaviorally as choosing to ask sexist questions during a mock job interview. As indicated above, sexist remarks are the most common examples of gender harassment found across many surveys.
Hypothesis 1: High sexist dyads (H1a), primed dyads (H1b), and dyads whose masculine identity has been threatened (H1c) are predicted to engage in more gender harassment during a mock job interview compared to their low sexist, non-primed, and non-threatened counterparts.
Hypothesis 2: High sexist dyads (H2a), primed dyads (H2b), and dyads whose masculine identity has been threatened (H2c) are predicted to evaluate a female interviewee more negatively after a mock job interview, compared to their low-sexist, non-primed, and non-threatened counterparts.
Hypothesis 3: An interaction between person and situation factors is predicted such that the relation between priming (Study 1) or masculine identity threat (Study 2) and gender harassing behavior will be stronger for male dyads with more sexist attitudes.
Hypothesis 4: An interaction between person and situation factors is predicted such that the relation between priming (Study 1) or masculine identity threat (Study 2) and negative evaluations of a female interviewee will be stronger for male dyads with more sexist attitudes.
Two hundred and eighty men in introductory psychology courses completed the Modern Sexism scale (Swim et al. 1995) together with several other self-report measures as part of a mass testing session. As part of the survey packet, respondents provided demographic information including their name and phone number. All participants in the mass testing session received partial extra credit in return for their participation. Participants were classified as either high or low in sexism by conducting a median split on the total sample of mass testing participants. Based on their classification as either high or low in Modern Sexism, men were contacted via phone about their willingness to participate in an additional research study.
Sexist attitudes were assessed during the mass testing session described earlier using the Modern Sexism scale (MS; Swim et al. 1995). The MS scale contains 13 items of which 8 are designed to assess modern sexism (i.e., denial of continuing discrimination, antagonism towards women’s demands, and resentment about special factors for women (e.g., It is easy to understand the anger of women’s groups in America.). The remaining 5 items measure more overt Old Fashioned sexism (e.g., Women are generally not as smart as men.). Respondents indicate their level of agreement/disagreement with a series of statements on 7-point scales from 1 (Strongly Disagree) to 7 (Strongly Agree). Dyad scores were created by averaging across each dyad member’s modern sexism score.
Interview question development and pretesting
One of the primary dependent measures concerned the types of questions selected to ask female interviewees during a mock job interview. Sixty-nine potential interview questions were developed to represent four distinct categories of interview questions: academic, research experience, miscellaneous, and personal/sexist questions. Each interview question was rated on an 11-point scale (0 = highly appropriate, 10 = highly inappropriate) by 21 college women regarding the appropriateness of each to ask during an interview situation. The average ratings for each of the question categories were: academic (M = 1.55, SD = .62), research experience (M = 1.42, SD = .66), miscellaneous (M = 1.51, SD = .56), and personal sexist (M = 8.57, SD = .46), and personal non-sexist (M = 2.33, SD = 1.09). Personal sexist questions were judged as significantly less appropriate (e.g., more sexist) to ask during an interview situation than questions from all of the other categories (ps < .001). Dyads were provided with eight questions from each of the aforementioned categories. For the personal category, 6 of the items were selected based on their perceived highly sexist nature and two questions were selected based on their personal but non-sexist nature.
Immediately after the first phase, which involved administration of the priming manipulation, participants completed a filler task purported to assess their attitudes toward censorship. The first item asked them to rate how sexually aroused they were by the film clip they had just viewed, followed by several filler questions regarding censorship. The final question again asked participants about their current level of sexual arousal. Both arousal measures were obtained on a 0–100 point scale with endpoints of 0 (Not at all aroused) to 100 (Extremely aroused).
Male dyad ratings of female interviewees
To assess the male dyads’ perceptions of the female interviewee, each dyad member rated the interviewee along several dimensions using 11 questions designed for purposes of the current research. Each question was scored on a 7-point scale with appropriately labeled endpoints. A principal components analysis (using varimax rotation) on average dyad ratings for each question, revealed three factors accounting for 70.35% of the variance. Eight items loaded on the first factor, termed “female competence”, accounting for 48.06% of the variance (e.g., In your opinion, how competent was the interviewee for the research assistant position? α = .89). Two items loaded on a second factor termed “female friendliness” which accounted for an additional 12.21% of the variance (e.g., How friendly did you find the interviewee? α = .76). A third factor, “female attractiveness” was determined by a single item accounting for an additional 10.09% of the variance (i.e., How attractive did you find the interviewee?). Composite scores were created for each factor by averaging across dyad scores; higher values indicated more competence, friendliness, and attractiveness. Questions used to evaluate the female were identical across studies 1 and 2. The reliability index for competency and friendliness in study 2 were .88 and .75, respectively.
Pairs of men were asked to participate in two short and purportedly unrelated studies: the first on censorship in the media and the second on the dynamics of job interviews. In the first phase, each dyad was randomly assigned to view either a 12 minute videotape sequence taken from segments of the commercially available film, Show Girls, or a 12 minute videotape sequence taken from a film produced by the National Endowment of the Arts (NEA) on censorship in the arts. The Show Girls videotape sequence was intended to prime thoughts about the sexist treatment of women. It was primarily a scene depicting a male producer graphically derogating the bodies of a group of young women who were auditioning for a chorus line. No nudity was included in the Show Girls segment. The NEA videotape sequence served as a control. It included a very academic discussion of how censorship has affected the arts.
Following each film, men were asked to rate on a 0–100 scale the degree to which they had been sexually aroused by the film. Subsequently, men answered 34 questions purported to measure their attitudes toward censorship. Next, they were asked to rate again their current sexual arousal level. While we believed that our priming manipulation could influence the degree to which men might be aroused, we also expected that answering a series of questions about censorship would allow time for any sexual arousal to dissipate. Measuring sexual arousal before and after the censorship questions allowed us to verify the expected reduction in arousal.
Following the “censorship study”, the two male participants walked down the hall to another room in which they expected to participate in a separate study on “job interviews” conducted by a separate experimenter. During this second phase, male dyads conducted a practice job interview with a female student whom they believed was applying for a job as a research assistant. Prior to the interview, dyads were asked to choose a series of interview questions from four question categories: Academic Experiences, Research Experience, Personal Information, and Miscellaneous. The questions were written on index cards and arranged in piles according to categories. A key dependent variable was derived from the dyads selection from the category entitled “Personal” questions. Two of the eight questions in this category were personal but non-sexist questions rated as innocuous by a pretest group of women as described earlier (e.g., If you had to describe yourself with one word, what would it be?). The other six questions, however, were selected to be maximally sexist and inappropriate questions for men to ask women in a job interview (e.g., How do you feel about wearing sexy clothing to work?). The number of these sexist questions selected by the interviewing duo constituted the primary behavioral measure of gender harassment in this study.
The men were instructed to interact when deciding which questions they would ask and to select eight questions in all. They were instructed to choose “some questions from each category.” Thus, it was possible for men to follow their instructions and still not select any of the sexist questions. They were given 4 min to make their selections while the experimenter left the room. Subsequently, the female interviewee (a confederate—five different women performed this role and their performances were counterbalanced across conditions) was brought into the interview room and seated across a table from the two male participants. Her responses were scripted. Some were simple responses to the questions (e.g., Q—On a scale from 1 to 10 how attractive would you rate yourself? A—I don’t know.). Others were more confrontational (e.g., Q—Do you think it’s important for women to wear bras to work? A—Of course it is, but I don’t see how that relates to being a research assistant.). Afterwards, the interviewee left while the two participants individually rated their perceptions of the interviewee. Finally, the men were debriefed and asked about their understanding of the procedures. None expressed suspicion about connections between the censorship and the interview studies.
Using Male Dyads as Units of Analyses
Following recommendations by Griffin and Gonzalez (1995), a series of correlations were computed to assess the dyadic interdependency of our male duos. First, a pairwise interclass correlation indicated a high level of similarity in sexism scores within dyads, r = .65, p < .001. Next, the overall pairwise correlations were computed to examine if the three female evaluation variables (competency, friendliness, and attractiveness) were related for the 126 individual men in Study 1. The significance test for this overall correlation depends on calculating an effective sample size. To this effect, we employed the procedure recommended by Griffin and Gonzalez for testing the significance of these relations. Significant relations emerged between competency and friendliness, r = .35, z = 3.93, p < .01, competence and attractiveness, r = .29, z = 3.22, p < .01, and female attractiveness and friendliness, r = .16, z = 2.00, p < .05.
Finally, we examined the extent to which these relations were driven by variability at the level of the dyad. To assess this, we computed a series of pairwise intraclass correlations, examining the extent to which dyad members were similar in how they evaluated the female interviewee across evaluative dimensions. That is, did dyad members resemble one another on ratings of the female interviewee’s competency, friendliness, and attractiveness? Significant dyad level variation emerged for attractiveness (pairwise intraclass correlation = .48, z = 3.81, p < .01). The pairwise intraclass correlations for competence and friendliness were not significant (ps > .05). Given the significant dyadic-level variability (in addition to individual-level variability noted above), both sources of variability are involved in how men evaluated the female interviewee. Given these findings, and consistent with the more general goal of using dyads (outlined above), we relied on dyads as the unit of analysis for all subsequent analyses.
Our priming manipulation sought to create differences in the degree to which male participants would view women in a sexist way. However, the film clip that we used as a sexist prime might also have produced difference in the degree to which our dyads were sexually aroused. We assessed dyad members’ sexual arousal immediately following the videotape and again after they completed the questionnaire on censorship. Our purpose in having participants complete the censorship questionnaire was to allow time for any sexual arousal created by the Show Girls film to dissipate. To assess the extent of sexual arousal at various time periods, dyadic levels of arousal were examined in a mixed-design ANOVA with one between-participants factor (Priming, 2-levels), one covariate (Modern Sexism) and one within-participants factor (Time-2 levels). This analysis revealed a Priming main effect, F(1, 59) = 14.42, p < .001, η2 = .20; a Time main effect, F(1, 59) = 20.28, p < .001, η2 = .26; and a Priming x Time interaction, F(1, 59) = 17.71, p < .001, η2 = .23.
Overall, Primed dyads indicated significantly greater sexual arousal (M = 21.78, SE = 1.65) compared to Non-Primed dyads (M = 1.06, SE = 1.52). In addition, dyads reported significantly higher levels of sexual arousal at Time 1, before completing the censorship questionnaire (M = 20.20, SE = 1.76) compared to after completing it (M = 2.73, SE = .72). The means related to the interaction showed that sexual arousal in the Primed condition did decrease after the censorship questionnaire, though a slight difference still remained between Primed and Non-Primed conditions after the questionnaire (before means: Primed: M = 38.61, SE = 2.59 vs. Non-Primed: M = 1.80, SE = 2.38 and after means: Primed: M = 4.94, SE = 1.06 vs. Non-Primed: M = .32, SE = .98).
To discount any interpretations of the subsequent analyses based upon the mediating effects of sexual arousal; later analyses were first performed with the Time 2 arousal levels as covariates. Arousal levels, however, were uncorrelated with any of the dependent variables and the inclusion of arousal levels as covariates did not alter the patterns of results.
Means, standard deviations, and zero-order correlations among variables for Study 1.
1. Modern Sexism
3. Sexist Questions
4. Female Competence
5. Female Friendliness
6. Female Attractiveness
The means, standard deviations, and zero-order correlations for Study 1 variables are presented in Table 1. Hypothesis 1a predicted that dyads with higher modern sexism scores would ask more sexist questions to a female interviewee during a job interview. In addition, hypothesis 1b predicted that Primed dyads would ask a greater number of sexist questions than Non-Primed dyads. To test these predictions, a general linear model ANOVA with one categorical factor (Priming, 2 levels) and one continuous variable (Modern Sexism) was used to predict the number of sexist questions asked by the male dyads. This analysis revealed a significant main effect of Priming, F(1, 59) = 5.82, p = .019, η2 = .09. Consistent with hypothesis 1b, Primed dyads asked significantly more sexist questions (M = 1.77, SE = .25) than Non-Primed dyads (M = 1.02, SE = .23). The predicted main effect for sexism failed to emerge. It was further predicted that an interaction would emerge between modern sexism and priming such that the relation between priming and gender harassment, would be stronger for individuals with higher sexist attitudes (Hypothesis 3). This interaction term approach a conventional level of significance, F(1, 59) = 3.48, p = .067, η2 = .06.
To better understand the nature of the interaction two methods were used: a simple slope analysis (Aiken and West 1991) and the Johnson-Neyman Technique, which provides a point estimate for the range of values of the moderator where the focal independent variable (priming) is significantly related to sexist question selection (Johnson and Neyman 1936; Hayes and Matthes in press). As illustrated in Table 3, and contrary to predictions, at high levels of sexism (+1SD), the relation between priming and sexist question selection was not significant, b = .08, t(58) = .17, p = .87. However, the simple slope was significantly different from zero at both the mean, b = .74, t(58) = 2.15, p = .035 and low (-1SD) levels of modern sexism, b = 1.40, t(58) = 2.88, p = .006. Moreover, the Johnson-Neyman point estimate indicated that when dyad modern sexism scores averaged 28.69 or above the coefficient for priming was significantly positive (using alpha = .05). Thus, dyads averaging high levels of modern sexism asked an equivalent number of sexist questions irrespective of whether they has previously been primed. In contrast, Primed dyads whose modern sexism scores were either average or below average, asked significantly more sexist questions compared to Non-Primed dyads. Thus, the behavior of less sexist dyads was influenced more by the priming manipulation than the behavior of more sexist men. This finding is discussed more below.
Ratings of the Female Interviewee
Hypothesis 2a predicted that dyads with higher sexism levels and primed dyads (Hypothesis 2b) would evaluate the female interviewee more negatively after a mock job interview. To test these predictions a general linear model MANOVA, using the 2-level Priming factor and a continuous variable (Modern Sexism), assessed female competence, attractiveness, and friendliness. Supporting hypothesis 2a, results revealed a highly significant multivariate main effect for Modern Sexism, F(3, 57) = 6.51, p = .001, Λ = .75. Follow-up univariate tests indicated the Modern Sexism main effect was driven by dyadic ratings of female competency, F(1, 59) = 11.91, p = .001, η2 = .17. Higher sexist dyads rated the female interviewee as less competent, overall, compared to dyads with lower sexism scores. Table 1 presents the correlations between Modern Sexism and confederate ratings. The predicted main effect for priming failed to emerge (Hypothesis 2b).
In support of hypothesis 4, which predicted an interaction between priming and sexism on dyad evaluations of the female interviewee, a borderline significant multivariate interaction emerged, F(3, 57) = 2.63, p = .059, Λ = .88. This interaction was driven by dyad ratings on female friendliness, F(1, 59) = 5.85, p = .019, η2 = .09; and competence, F(1, 59) = 5.15, p = .027, η2 = .08. As illustrated in Table 3, at low and mean levels of sexism, both Primed and Non-Primed dyads rated the female as equally competent. However, high sexist dyads tended to rate the interviewee as less competent after being primed, b = −.41, t(58) = −1.88, p = .063. The Johnson-Neyman point estimate for competency was 38.36, indicating a significant effect of priming on competency for dyads whose average sexism score was at this level or higher. A similar pattern emerged for friendliness such that the coefficient for the effect of priming on friendliness was strongest for high sexist dyads, b = −.39, t(58) = −1.93, p = .058. In addition, the Johnson-Neyman point estimates indicated that priming had a significant positive effect on ratings of female attractiveness (increasing attractiveness ratings) when the dyads’ modern sexism scores averaged 13.37 or lower; in contrast, priming had a significant negative effect on ratings of female friendliness when dyad average sexism scores were 37.29 or higher.
Discussion of Study 1
The results of Study 1 indicated that both person and situation factors are important to the prediction of gender harassment. As predicted, men who were primed with a video showed women portrayed in sex-stereotypic terms were more likely to ask sexist questions during a subsequent mock job interview than those who were not primed.
However, overall, higher sexist dyads did not ask more sexist questions of the female interviewee than lower sexist dyads. Yet, high sexist dyads were more likely to rate the female interviewee’s competence lower than low sexist dyads, r(63) = −.35, p = .005. Thus, dyads who reported more sexist attitudes, on average, seemed more inclined to manifest their sexism via more covert cognitive routes as compared to more overt behavioral routes. Although not predicted, this finding is not surprising given many of the correlates of the Modern Sexism scale relate to the cognitive derogation of women.
The obtained interactions between sexism and priming on gender harassment and dyad evaluations of the female interviewee again attest to the importance of both personal and situational variables in developing and/or maintaining negative response toward women. Nevertheless, the direction of some of these effects is not entirely consistent with predictions. For example, the finding that the question selection of low sexist primed dyads was most influenced by the priming manipulation may indicate that, while these men are not apt to hold negative attitudes toward women, sexual priming may serve to elicit otherwise non-normative behavior. Additional interpretations of this effect are discussed more in the general discussion section below.
Study 2 was conducted to further assess the range of situational variables that may influence gender harassing behavior. As mentioned earlier, one factor thought to affect the likelihood of men engaging in gender harassment is gender identity threat. For this study, identity threat was manipulated by telling half of the participants that they had been out-performed by a female on a task where men were expected to out-perform women. Similar to Study 1, following the manipulation of threat, male dyads with similar levels of sexist attitudes evaluated the female in an interview for a research assistant position. It was predicted that more highly sexist and threatened dyads would select a greater number of sexist questions to ask the female during the interview and rate the female more negatively following the interview than lower sexist (Hypothesis 1a) and non-threatened dyads (Hypothesis 1c), respectively. Consistent with the Person X Situation approach, we also examined the interaction between sexism and identity threat in predicting behavioral (Hypothesis 3) and cognitive sexism (Hypothesis 4).
Similar to Study 1, 188 men were first pre-tested using the Modern Sexism Scale during a mass testing session. The procedure for selecting men for participation was identical to that discussed for Study 1 above.
Masculine Knowledge Test. Fifty-eight multiple-choice questions were generated by the researchers with assistance from two undergraduate women. The topics of included sports, automobile mechanics, alcohol, and home improvement—all selected to be stereotypically masculine interests. Sixteen undergraduate men were asked to rate each multiple choice question using two scales: “How likely is it that the average college male will think that he knows the correct answer to this question?” and “How likely is it that the average college male will assume that women typically do not know the answer to this question?” Both ratings were made on 7-point scales with the anchors 1 (Not at all likely) to 7 (Very likely). Ratings were averaged for each multiple choice question, and the 25 most highly rated questions were selected to form the “Masculine Knowledge Test.” Average scores for these 25 questions ranged from 5.02 to 5.75.
Pairs of men were asked to participate in two short and purportedly unrelated studies: the first on test development and the second on the dynamics of job interviews. In the first study, each dyad was randomly assigned to either a Masculinity Threat condition or a Non-Threatening condition. In the Masculinity Threat condition, each male dyad participated along with a female confederate (5 confederates were used, counterbalanced across conditions). The participants were led to believe that the confederate was just another participant.
Following the completion of these tests, the experimenter collected them and then retreated to another room purportedly to score them. He then returned and gave the three participants feedback on their performances. In the Masculinity Threat conditions he said to the woman, “Your results are sort of surprising. The Principle investigator has given this test to hundreds of ISU students and you have not only performed significantly better than the vast majority of women on the test, your performance is also significantly higher than most men.” Then the two men were told that they had scored slightly below the average for college men. The procedures in the Non-Threatening conditions were identical to those in the Masculinity Threat conditions except for the feedback. In the Non-Threatening conditions, the woman was told that she had scored at an average level for women. It was emphasized that women typically do not score as high as men on the test. So, her score was below those usually found for men. The men were told that they had both scored slightly above average for college men.
Next, the participants and confederate were asked to walk down the hall to another room in which they expected to participate in a separate study. The procedure for this phase was identical to those used in the job interview phase of Study 1. A rigged drawing was used to make it appear that the two men had been randomly selected to play the roles of the interviewers and that the woman had been randomly selected to play the role of the interviewee.
Using Male Dyads as Units of Analyses
Similar to Study 1, a series of correlations were computed to assess the dyadic interdependency. Once again, the pairwise interclass correlation indicated a high level of similarity in sexism scores within dyads, r = .73, p < .001. Next, the overall pairwise correlations were computed to examine if competency, friendliness, and attractiveness were related for the 104 individual men in Study 2. Significant relations emerged between competency and friendliness, r = .49, z = 4.95, p < .01, competence and attractiveness, r = .23, z = 2.30, p < .05, and attractiveness and friendliness, r = .19, z = 2.00, p < .05. Consistent with Study 1, significant dyad level variation emerged for attractiveness (pairwise intraclass correlation = .43, z = 3.10, p < .01) meaning both dyad members rated the female similar in attractiveness. The pairwise intraclass correlations for competence and friendliness were not significant (ps > .05) meaning that dyad members ratings on these dimensions were relatively independent of one another. Given this partial evidence of dyad level variability and to be consistent with Study 1, dyads were selected to represent the unit of analysis.
Means, standard deviations, and zero-order correlations among variables for Study 2.
1. Modern Sexism
2. Identity Threat
3. Sexist Questions
4. Female Competence
5. Female Friendliness
6. Female Attractiveness
The means, standard deviations, and zero-order correlations for those variables examined within Study 2 are presented in Table 2. Hypothesis 1a predicted that dyads with higher modern sexism scores would ask more sexist questions of a female interviewee during a mock interview task. It was also predicted that dyads whose masculinity had been threatened would ask more sexist questions than non-threatened dyads (Hypotheses 1c). These hypotheses received partial support. A general linear model ANOVA with one categorical factor (Masculinity Threat, 2 levels) and one continuous variable (Modern Sexism) was used to predict the number of sexist questions asked by the male dyads. This analysis revealed a significant main effect for Masculinity Threat, F(1, 45) = 5.06, p = .029, η2 = .10. Consistent with hypothesis 1c, men in the Masculinity Threat condition asked significantly more sexist questions (M = .75, SE = .16) than those in the Non-Masculinity Threat condition (M = .25, SE = .16). The predicted group sexism main effect (Hypothesis 1a) and the interaction between sexism and threat (Hypothesis 3) failed to emerge.
Ratings of the Female Interviewee
It was predicted that higher sexism dyads and dyads whose masculinity had been threatened would rate the female interviewee more negatively after a mock job interview (hypothesis 2a and 2c, respectively). A multivariate general linear model with one categorical factor (Masculinity Threat, 2 levels) and one continuous predictor (Modern Sexism) was computed. Supporting hypothesis 2c, the predicted multivariate main effect for Masculinity Threat was significant, F(3, 42) = 4.37, p = .009, Λ = .76. Follow-up univariate tests indicated that this effect was driven by effects on female competency, F(1, 44) = 8.60, p = .005, η2 = .16, and to a lesser degree on female attractiveness, F(1, 44) = 3.08, p = .086, η2 = .07. Dyads in the Masculinity Threat condition rated the female interviewee as less competent (M = 5.45, SE = .10) and attractive (M = 2.55; SE = .23) compared to Non-Threatened dyads (competence: M = 5.53, SE = .10; attractiveness: M = 3.47, SE = .23). The predicted main effects of Modern Sexism on evaluations of the female interviewee failed to emerge (Hypothesis 2a).
Simple slope tests and point estimates for the interaction between priming (Study 1) and masculine identity threat (Study 2) and sexist attitudes on the dependent measures.
Dyad Sexism Level
Sexist Question Selection
Low Sexism (-1SD)
High Sexism (+1SD)
The two studies reported here found evidence that both person and situation factors are related to the likelihood of men engaging in gender harassing behavior. More specifically, Study 1 examined one person factor, sexist attitudes, as measured via the Modern Sexism scale (Swim et al. 1995) and one situational factor where male dyads were either primed or not to view women in sex-stereotypic terms. We defined gender harassment behavior as the number of sexist questions men posed to women during a mock job interview as well as male dyads’ impressions of the female interviewee. As predicted, more sexist questions were asked by dyads of men who were previously primed with sexist media (Study 1) and by dyads whose sense of masculinity identity was threatened (Study 2). In addition, although the predicted main effect for sexist attitudes on sexist question selection during a mock job interview was not supported, sexist attitudes (a person factor) did interact with situational factors in eliciting gender harassing behavior and negative evaluations of the female interviewees.
More specifically, in Study 1, the interaction between sexism and priming on sexist question selection was not exactly as predicted. We expected that, for dyads low in sexism, priming would have little impact on sexist question selection. This prediction was based on the assumption that, overall, these men are less likely to hold a variety of attitudes associated with viewing men and women in a “we/they” framework across situations. As such, these men should be more constrained in their behavior even if a situation somehow encouraged sexist behavior. Consistent with this, we expected little effect of priming on the behavior and attitudes of low sexist men (dyads); however, we expected high sexist men to be less constrained in their behavior and more reactive to situational influences.
While interactions did emerge between sexism and priming for our behavioral measure of gender harassment, the results were not entirely consistent with predictions. More specifically, the priming manipulation had an impact upon the number of sexist questions selected by low sexist dyads, but no impact upon question selection in high sexist dyads. This pattern might suggest that highly sexist men need no provocation to behave in a sexist manner, but low sexist men do. Perhaps a sexist schema is chronically accessible for highly sexist men, but can be made temporarily accessible for low sexist men when they are primed (Higgins 1996). Nevertheless, high sexist dyads were found to evaluate female interviewees as less competent, especially after being sexually primed. Thus, in addition to the direct effect of offending women who are exposed to sexist media at work, such stimuli possess the power to affect women indirectly in the workplace by making their male coworkers more likely to exhibit sexist behaviors and evaluate them more negatively.
If one considers the fact that the Modern Sexism scale is based on the earlier concept of Modern Racism, one can begin to develop a theoretical framework underlying these findings. For example, the MS scale is not about sexual objectification but rather the notion that women are too pushy and don’t deserve special treatment. Viewed from this perspective, it makes sense that the MS scale would not relate directly to sexist question selection as the questions themselves were more indicative of traditional sexist beliefs about women (i.e., wearing a bra, perfume, or sexy clothes to work). However, because of its focus on women’s role in the “modern world”, it would be expected to affect how qualified and competent a women applicant would appear for a research assistant position. To more fully examine this proposition, we examined if dyads’ Old Fashion Sexism (OFS) scores were a stronger predictor of gender harassing behavior than dyad Modern Sexism scores in Study 1 (due to practical limitations involving the length of the mass testing session, OFS was not assessed in Study 2). Although MS scores were not directly related to gender harassment in Study 1, r(61) = .10, ns, average dyad OFS scores were significantly related to gender harassment, r(61) = .20, p = .024. Our interpretation of this effect is further supported by Swim et al.’s research on sexism. In contrast, the priming manipulation is likely to have increased the accessibility of sex-related concepts and resulted in a greater likelihood of choosing questions with sexual content (Bargh 1990).
As previously mentioned, in Study 2 the situational variable of masculine identity threat was also related to the number of sexist questions selected by male dyads. Females who outperformed males on a “masculinity test” were asked significantly more sexist questions, on average, than when male dyads were told they had outperformed a female on a masculinity test. The obtained interactions between sexism and masculinity threat on interviewee competency and attractiveness also support predictions such that the masculinity threat manipulation resulted in lower competency and attractiveness evaluations of the female interviewee, but only for high sexist dyads. Thus, these findings support the personality constraint position advocated above and are consistent with the Person X Situation model advanced by Pryor and Whalen (1997).
At present, we know of only one other study that has specifically examined a connection between threats to masculinity and gender harassment (Maass et al. 2003). A higher incidence of sexual harassment in “male culture” work environments is well documented (Gruber 1998). However, inferences about a threat to masculine identity as a motivation for gender harassment in “male culture” jobs have come largely from case studies (e.g., Martin 1990). One might argue that all forms of harassment are simply more common in so-called “male culture” environments and that women who enter these environments are simply entering the fray to which men are subjected all the time (see Faludi 1994). In this context, the results of Study 2 establish that a woman outperforming men on a masculine task is at least a sufficient cause for gender harassment.
Previous research has documented that women who cross the boundaries of traditional gender roles to enter “a man’s world” are often subjected to sexual harassment from their male coworkers (Brown 1998). For example, the MSPB survey (1995) found that women were more likely to report experiencing sexually harassing behavior when their co-workers were all men. Also, Bastian et al. (1996) reported that almost a third of military women who experienced sexually harassing behavior indicated that it happened in a work environment where personnel of their gender were uncommon. Pryor and Whalen (1997) have suggested that gender harassment often represents men’s expressions of hostility toward women perceived as “outsiders” or members of an out-group. Thus, when the opportunity arises, in-group members tend to devalue the performance of out-group members especially when an out-group member performs better than the in-group (Schmader and Major 1999).
The present research tends to suggest that such devaluation may manifest itself in different ways. For example, in study 2 it would not make logical sense for the men to report that a female who just performed better than them on a “masculine” task as being incompetent. Yet, these men were willing to devalue the female in areas not directly related to the task being performed (e.g., attractiveness). Turner et al. (1987) maintain that in-group favoritism is the product of people defining themselves and their self-esteem in relation to their in-group associates. In-group/out-group categorizations may automatically activate evaluative structures that can influence expectations about and perceptions of group members (Dovidio and Gaertner 1993).
The strongest evidence for the Person X Situation model (Pryor and Whalen 1997) came from analyses of men’s ratings of the competence of the women they interviewed in the mock job interviews. In Study 1, the lowest ratings of female interviewees’ competence came from highly sexist men who were exposed to sexist primes. In Study 2, the lowest competence ratings came from highly sexist men whose masculinity identity was threatened by being bested by a woman on a masculine knowledge test. When male participants were asked to interview and evaluate a woman for a job, those who were predisposed toward sexism and who were exposed to situational factors facilitating gender harassment were most likely to devalue the woman’s competence for performing the job.
Gender harassment is one of several possible forms of sexually harassing behavior. Surveys have consistently indicated that gender harassment from one’s peers or coworkers is by far the most common form of sexual harassment. One of the immediate effects of this kind of behavior maybe to undermine women’s sense of self-confidence, which might be particularly so for women who buy into gender role stereotypes. Consistent with this analysis, Swim and Hyers (1999) found that sexist comments had a negative impact upon women’s state self-esteem for those who were traditional in their gender role orientations. Clearly, the continuing prevalence of gender harassment in the workplace represents a significant barrier to women’s achievement of equal opportunities in employment.
The current research was limited to examining instances of gender harassment among a sample of undergraduate men within the U.S. and, as a result, future research using controlled laboratory methods like those used in the current studies along with field research could further explore variables that might reduce the incidence of gender harassment. Another direction for future research would be to examine the replicability of these findings using a cross-cultural perspective to help establish the generalizability of the proposed model of gender harassment. It also seems valuable to conduct research examining factors that negate sexist priming, that induce men to feel less threatened by women who outperform them, that ameliorate the harassing proclivities of highly sexist men, or that enhance the positive influence of less sexist men upon a group’s refraining from harassing behavior.
The authors wish to thank Ana Couri, Brooke Daitchman, Jennifer DeStefano, Annette Edwards, Mandy Erickson, Jennifer Eberle, Jaimee Foster, Amanda Goldstein, Eric Hart, Kerry Herbert, Julie Hoffman, Ravi Kalani, Tracey Little, Stephanie Meyer, Michelle Morse, Shana Peterson, Bonnie Schieber, Scott Seely, Valerie Snell, Emannuela Sprizzi, Ann Szpytek, Wolfgang Viechtbauer, Michelle Weiss, and Jackie Woods for their able assistance in conducting these studies.