WORKING WOMEN

The Sexual Harassment of Uppity Women

Jennifer L. Berdahl University of Toronto

In 3 studies, the author tested 2 competing views of sexual harassment: (a) It is motivated primarily by sexual desire and, therefore, is directed at women who meet feminine ideals, and (b) it is motivated primarily by a desire to punish gender-role deviants and, therefore, is directed at women who violate feminine ideals. Study 1 included male and female college students (N � 175) and showed that women with relatively masculine personalities (e.g., assertive, dominant, and independent) experienced the most sexual harassment. Study 2 (N � 134) showed that this effect was not because women with relatively masculine personalities were more likely than others to negatively evaluate potentially harassing scenarios. Study 3 included male and female employees at 5 organizations (N � 238) and showed that women in male-dominated organizations were harassed more than women in female-dominated organi- zations, and that women in male-dominated organizations who had relatively masculine personalities were sexually harassed the most.

Keywords: gender, sex, sexual harassment, stereotypes, women

Uppity: Taking liberties or assuming airs beyond one’s station; pre- sumptuous: “was getting a little uppity and needed to be slapped down” (New York Times). (American Heritage Dictionary of the English Language, 2004)

The original prototype of sexual harassment was a male boss sexually coercing a female subordinate. Sexual harassers were assumed to be motivated by sexual desire for their targets (cf. Franke, 1997; Gutek, 1985; MacKinnon, 1979; Schultz, 1998). If sexual harassment is motivated by sexual desire, then the most frequent targets of sexual harassment should be individuals who meet gender ideals. Gender ideals involve both physical and per- sonality characteristics. Personality characteristics desired in men include assertiveness, independence, and dominance; those desired in women include modesty, deference, and warmth (Bem, 1974; Connell, 1987; Eagly, 1987; Fiske & Stevens, 1993; Prentice & Carranza, 2002). If women are sexually harassed more than men, and if individuals who meet gender ideals are harassed more than those who do not, then women with feminine personalities should be sexually harassed the most.

I suggest that just the opposite is true. The most common form of sexual harassment is gender harassment (Fitzgerald, Shullman, et al., 1988), a form of hostile environment harassment that ap- pears to be motivated by hostility toward individuals who violate gender ideals rather than by desire for those who meet them (e.g., Berdahl, Magley, & Waldo, 1996; Dall’Ara & Maass, 1999; Fiske

& Stevens, 1993; Franke, 1997; Maass, Cadinu, Guarnieri, & Grasselli, 2003). Whether women who violate feminine ideals experience more sexual harassment generally, including unwanted sexual attention and coercion, is not known. In this article, I propose that they do. I conducted three studies to test the predic- tion that women who violate feminine ideals are most likely to be sexually harassed in ways traditionally identified as harassing to women. In Study 1, I examined how young adults’ experiences of sexual harassment across a variety of life domains relate to their sex and personality gender (masculinity and femininity). In Study 2, I tested whether women with masculine personalities are more likely than others to negatively evaluate potentially harassing experiences. In Study 3, I examined how employees’ experiences of sexual harassment in five organizations relate to their sex, occupational gender (male-dominated or female-dominated), and personality gender.

This set of studies helps to establish whether sexual harassment in general is targeted more at women who meet gender ideals or at those who violate them. By implication, this sheds light on whether sexual harassment as traditionally construed, not just gender harassment, is motivated by sexual desire or by sexist antipathy. By focusing on personality gender violations, I exam- ined a relatively subtle form of violating gender ideals and offer a strict test of whether sexual harassment is primarily targeted at “uppity” women who step out of place by assuming characteristics considered more desirable for men. This research also represents the first to study systematically how women’s experiences of sexual harassment relate to their gender role conformity in their real lives and places of employment.

Sexual Harassment: Aimed at the Desirable or the Deviant?

At the Desirable

Sexual harassment is widely assumed to stem from a desire for sexual expression or gratification and, by implication, to be tar-

This research was financially supported by the Connaught Foundation and Petro Canada. I thank members of the audience at an early presentation of some of this research to the Psychology Department at Princeton University in October 2003. I am indebted to Susan T. Fiske for her insightful feedback on an earlier version of this article.

Correspondence concerning this article should be addressed to Jennifer L. Berdahl, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada, M5S 3E6. E-mail: [email protected]

Journal of Applied Psychology Copyright 2007 by the American Psychological Association 2007, Vol. 92, No. 2, 425– 437 0021-9010/07/$12.00 DOI: 10.1037/0021-9010.92.2.425

425

geted at those who arouse this desire. The first type of sexual harassment to be identified was quid pro quo harassment from a male boss toward a female subordinate (Williams v. Bell, 1978). Courts originally assumed that harassers were motivated by sexual desire for their targets and reasoned that sexual harassment was sex discrimination because sexual acts are necessarily motivated by the target’s sex (see Franke, 1997, for a review). Theories of sexual harassment shared this assumption (for reviews, see Lengnick-Hall, 1995; Tangri & Hayes, 1997; Welsh, 1999). The natural– biological model proposed that sexual harassment results from natural and inevitable feelings of sexual desire expressed primarily by men toward women because of inherent sex drives and functions (cf. Studd & Gattiker, 1991). Power models viewed sexual harassment as resulting from the fact that men’s economic power over women enables them to exploit and coerce women sexually (Equal Employment Opportunity Commission [EEOC], 1980; Evans, 1978; Farley, 1978; MacKinnon, 1979; Nieva & Gutek, 1981; Zalk, 1990). Sex role spillover theory proposed that sexual harassment is a form of sociosexual behavior at work that is guided by the roles of men as sexual agents and women as sexual objects (Gutek, 1985; Gutek, Cohen, & Konrad, 1990; Gutek & Morasch, 1982).

If sexual harassment is motivated by sexual desire, then indi- viduals who elicit this desire should be the primary targets of harassment. Gender ideals capture what is considered most desir- able in men and women and include both physical and personality characteristics. Personality characteristics considered most desir- able in men include assertiveness, independence, and dominance; those considered most desirable in women include modesty, def- erence, and warmth (Bem, 1974; Connell, 1987; Eagly, 1987; Fiske & Stevens, 1993; Prentice & Carranza, 2002). If mostly men harass mostly women because of sex differences in biology, power, or social roles, then women who meet feminine ideals, including feminine ideals for personality, should be sexually ha- rassed the most.

At the Deviant

After the initial focus on quid pro quo sexual harassment, other forms of sexual harassment were identified. Hostile environment harassment does not involve trying to establish sexual or romantic relations with a target but, rather, makes a target feel unwelcome in the workplace on the basis of sex (Harris v. Forklift Systems, 1993; Meritor Savings Bank v. Vinson, 1986). Also called gender harassment (Fitzgerald, Shullman, et al., 1988) or sexist or sexual hostility (Fitzgerald, Magley, Drasgow, & Waldo, 1999), this is the most common type of sexual harassment (e.g., Fitzgerald, Shull- man, et al., 1988; Fitzgerald, Drasgow, Hulin, Gelfand, & Magley, 1997; Fitzgerald et al., 1999; Franke, 1997; Schultz, 1998; U.S. Merit Systems Protection Board, 1981, 1988, 1995; Waldo, Ber- dahl, & Fitzgerald, 1998). Gender harassment undermines, humil- iates, or rejects a target on the basis of sex with sexual and sexist remarks, jokes, materials, or pranks. Gender harassment appears to be motivated by sexist hostility, not sexual desire (e.g., Fiske & Glick, 1995; Franke, 1997; Maass, Cadinu, Guarnieri, & Grasselli, 2003; O’Leary-Kelly, Paetzold, & Griffin, 2000; Schultz, 1998; Maass et al., 2003; Stockdale, Visio, & Batra, 1999).

Evidence suggests that gender harassment against women is primarily targeted at those who violate gender ideals (Burgess &

Borgida, 1999; Dall’Ara & Maass, 1999; Fiske & Stevens, 1993; Franke, 1997; Maass et al., 2003). One example includes an assertive female police officer and bodybuilder who was subjected to sexually explicit noises and materials and found vibrators, a urinal device, and a soiled condom and sanitary napkin in her mailbox at work (Sanchez v. City of Miami Beach, 1989). Another includes the famous case of Price Waterhouse v. Hopkins (1989), in which an outspoken and extremely high-performing woman in a male-dominated professional accounting office was denied part- nership and instructed to learn how to “walk more femininely, talk more femininely, dress more femininely, wear make-up, have her hair styled, and wear jewelry” (p. 235). This case led some to propose that this kind of sex discrimination is motivated by a desire to punish women who do not conform to prescriptive sex stereotypes or to beliefs about how women should behave (Bur- gess & Borgida, 1999; Fiske & Stevens, 1993).

Recent experiments provide compelling evidence that this is the case. Using a computer paradigm, Maass and colleagues had men receive an electronic communication from a purported interaction partner (Dall’Ara & Maass, 1999; Maass et al., 2003). Half of the men received a message from a woman who said she was studying economics, intended to become a bank manager, thought women were as capable as men, and participated in a union that defended women’s rights. The other half of the men received a message from a woman who said she was studying education, intended to become an elementary school teacher to allow time for family and children, and chose not to become a lawyer because the job is more appropriate for men and she is afraid to compete with men. Men had the option of sending a variety of images to their interaction partner in reply and were more likely to send offensive pornogra- phy to the woman who expressed nontraditional beliefs and career ambitions than to the woman who expressed traditional ones.

The rationale provided by Maass et al. (2003) for why men gender harass nontraditional women is that men are motivated to derogate women when they experience a threat to their male identity. Women threaten male identity when they blur distinctions between men and women and thereby challenge the legitimacy of these distinctions and the status they confer men. The fictitious women in the computer paradigm experiments represent archetypi- cal cases of gender role conformists and violators. Men who endorsed gender role distinctions and who were strongly male identified were more likely than other men to gender harass the nontraditional woman and to report an enhanced sense of male identity after doing so (Dall’Ara & Maass, 1999; Maass et al., 2003).

Sexual approach forms of harassment derogate women as well. Sexually objectifying a woman and treating her as a means to an end derogates her, especially in the sociohistorical context of sexual exploitation and violence against women (e.g., Koss et al., 1994; Schacht & Atchison, 1993; Schultz, 1998). Unwanted sexual attention, repeated requests for dates, and sexual coercion are ways of derogating women in addition to gender harassment. Viewed in this way, it is not surprising that gender and sexual approach forms of harassment are highly empirically related (e.g., Fitzgerald, Gel- fand, & Drasgow, 1995). Given this, it seems reasonable to pro- pose that they share a common motive. After all, men with rela- tively sexist attitudes are not only more likely to gender harass women (Dall’Ara & Maass, 1999; Maass et al., 2003) but they are more likely to sexually exploit them as well (Pryor, 1987; Pryor,

426 BERDAHL

LaVite, & Stoller, 1993). Sexual approach forms of harassment may appear to be sexually motivated, but they may be more basically motivated by a desire to derogate the target of harass- ment, especially a woman who steps out of “place.” An extreme example of this is a woman who successfully passed as a man and was raped by two male “friends” when they learned she was female (Minkowitz, 1994).

I propose that sexual harassment generally derogates the target of harassment on the basis of sex and is most frequently targeted at women who violate prescriptive gender roles (Berdahl, in press):

Hypothesis 1: Women who have characteristics or engage in behavior considered more desirable for men than for women experience more sexual harassment than other women and men.

Consistent with this hypothesis, research has shown that women in male-dominated occupations, especially those in male- dominated work contexts, are sexually harassed more than women in balanced or in female-dominated ones (Fitzgerald, Drasgow, Hulin, Gelfand, & Magley, 1997; Glomb, Munson, Hulin, Berg- man, & Drasgow, 1999; Gruber, 1998; Mansfield et al., 1991). This would also be the case, however, if men are more sexually aggressive than women, sexual harassment is driven by sexual desire, and women who meet gender ideals are sexually harassed the most. To test whether sexual harassment is targeted more at women who violate gender ideals, it is necessary to compare women who differ in their conformity to these ideals in the same context. Maass and colleagues (Dall’Ara & Maass, 1999; Maass et al., 2003) began this with their experiments showing that a ficti- tious woman representing a feminist archetype is more likely than one representing a traditional archetype to be targeted for a spe- cific act of gender harassment. The current set of studies tested whether sexual harassment generally is targeted at women who violate gender ideals and whether it is targeted at women who violate gender ideals in more subtle ways than their expressed career ambitions and attitudes about women’s rights. I examined sexual harassment as traditionally construed to include sexual approach forms of harassment, such as unwanted sexual attention and sexual coercion, and studied how actual women’s experiences of sexual harassment in their lives and places of employment relate to women’s personality gender violations.

Gender ideals, or beliefs about how men and women should think, feel, and behave, are captured by prescriptive sex stereo- types (Eagly & Karau, 2002; Fiske & Stevens, 1993; Heilman, 2001; Rudman & Glick, 1999, 2001). These are likely to differ somewhat across contexts (cf. Alvesson & Billing, 1997; Connell, 1987; Prentice & Miller, 2002), but there is much commonality in feminine and masculine ideals across a wide range of environ- ments (Bergen & Williams, 1991; Buss, 1989; Connell, 1995; Eagly, Wood, & Diekman, 2000; Fiske, Cuddy, Glick, & Xu, 2002; J. E. Williams & Best, 1990). The best-known and most widely used measure of gender ideals for personality is the Bem Sex Role Inventory (BSRI; Bem, 1974). The BSRI was created by identifying personality characteristics considered more desirable for men than for women and characteristics considered more desirable for women than for men in American society. The former characteristics capture “masculine” ideals and include acting as a leader and being assertive, ambitious, dominant, and independent.

The latter characteristics capture “feminine” ideals and include being affectionate, compassionate, gentle, warm, and sensitive to the needs of others. Despite being developed more than 30 years ago, recent studies have shown that the BSRI continues to capture masculine and feminine ideals in American society (cf. Holt & Ellis, 1998; Prentice & Carranza, 2002).

In the current studies, I used the BSRI to test the prediction that women with relatively masculine personalities are more likely than other women (and men) to experience sexual harassment. I began with a study of upper level undergraduates to test this prediction because sexual harassment occurs in a variety of domains, includ- ing school (e.g., Cortina, Swan, Fitzgerald, & Waldo, 1998; Duffy, Wareham, & Walsh, 2004; Fitzgerald, Shullman, et al., 1988; McCormack, 1985; Paludi, 1997; Reilly, Lott, & Gallogly, 1986). In a second study, I tested whether women who have relatively masculine personalities are more likely than others to negatively evaluate the same potentially harassing behaviors to see whether this might explain the results of Study 1. In Study 3, I tested whether the results for Study 1 generalize to workplace settings by examining the sexual harassment experiences of employees at work.

Study 1

Sample

Participants were 175 students (77 men) enrolled in undergrad- uate management courses at a large public university in the North- east who participated in exchange for course credit. They came to a large classroom in groups of approximately 20 and individually completed questionnaires. Sixty-seven percent identified their eth- nic background as Asian; 25% as Caucasian; 3% as Arab or Middle Eastern; 2% as Caribbean; and 1% or less as African, Hispanic, or other. Average age was 20.67 years (SD � 2.01).

Measures

Personality gender. Personality gender was measured with the short form of the BSRI (Bem, 1978). This measure includes 10 masculine characteristics rated as more desirable for men than for women in American society, 10 feminine characteristics rated as more desirable for women than for men, and 10 neutral character- istics rated as equally desirable for men and for women. The short form of the BSRI is more reliable than the long form, which contains twice as many items (Campbell, Gillaspy, & Thompson, 1997). Importantly, the short form omits the terms masculine and feminine, which may alert respondents to what is being measured.

The sample consisted of a majority of students who identified their ethnic background as Asian, consistent with the student population from which they came. The BSRI was developed primarily with Caucasian students (Bem, 1974). To make sure that the population being studied viewed the characteristics on the BSRI as representing gender ideals in society, I had a separate sample of 46 Caucasian and 85 Asian students from the same population rate how desirable it is in American society for a man or a woman to have each characteristic on the BSRI (from 1 � not at all desirable to 7 � extremely desirable, consistent with Bem’s original instructions; Bem, 1974). Asian and Caucasian students rated the masculine items as significantly more desirable for men

427SEXUAL HARASSMENT

than for women (see Table 1); Caucasians: F(1, 45) � 21.37, p � .0001, d � 1.37; Asians: F(1, 83) � 22.60, p � .0001, d � 1.02. They rated the feminine items as significantly more desirable for women than for men; Caucasians: F(1, 45) � 30.59, p � .0001, d � 1.64; Asians: F(1, 83) � 14.71, p � .0001, d � .84. Asians did not differ from Caucasians in how desirable they rated the masculine items for a man, F(1, 60) � .27, ns, the feminine items for a man, F(1, 60) � 3.87, ns, the masculine items for a woman, F(1, 71) � 3.55, ns, or the feminine items for a woman, F(1, 70) � .31, ns. This suggests that the BSRI is an appropriate measure of societal gender ideals for personality for this population.

Respondents for the sexual harassment survey rated how true of them each of the 30 characteristics on the short form of the BSRI was on a scale ranging from 1 (never or almost never true) to 7 (always or almost always true). Alpha reliabilities for the mascu- line (� � .82) and feminine (� � .86) characteristics were good. Individuals’ self-ratings on the masculine items were averaged to compute individual masculinity scores, and their self-ratings on the feminine items were averaged to compute individual femininity scores. Consistent with scoring methods recommended for the BSRI (Bem, 1978), I classified individuals as masculine, androg- ynous, feminine, or undifferentiated depending on whether their scores were higher or lower than median scores. Because assess- ments of masculinity and femininity tend to be made within sex (e.g., Connell, 1987; Locksley & Colton, 1979)—for example, women assess their warmth compared with other women and men assess their aggressiveness compared with other men—I used within-sex medians to classify individuals (male medians: mascu- linity � 4.60, femininity � 4.80; female medians: masculinity � 4.40, femininity � 5.20). Individuals were classified as masculine if their masculinity score exceeded the masculinity median and their femininity score was equal to or lower than the femininity median; individuals were classified as androgynous if their scores exceeded both medians; individuals were classified as feminine if their masculinity score was equal to or lower than the masculine median and their femininity score exceeded the feminine median; individuals were classified as undifferentiated if their scores were equal to or lower than both medians (see Table 4 for the number of respondents in each category).

Sexual harassment. Sexual harassment was measured with the same 14 items used by Berdahl and Moore (2006) to measure traditional forms of sexual harassment. These items were derived from the Sexual Experiences Questionnaire (Fitzgerald, Shullman, et al., 1988; Fitzgerald et al., 1995), and, like the Sexual Experi- ences Questionnaire, were written in specific behavioral terms so that respondents did not have to judge whether they were harassed

but simply had to indicate how often they had a particular expe- rience (from 0 � never to 4 � most of the time).1 Respondents indicated how often they had each experience in the past 2 years at school, work, or among family and friends. As full-time under- graduates, some did not have enough work experience to limit the question to work, and sexually harassing behaviors occur in a variety of life domains, including school (e.g., Cortina et al., 1998; Duffy et al., 2004; McCormack, 1985; Reilly et al., 1986).

The items describe experiences that both men and women can have (e.g., “Made sexist comments or jokes,” “Attempted to es- tablish a romantic or sexual relationship despite your efforts to discourage it,” and “Treated you badly for refusing to have sexual relations with them”) but were originally written to capture sexual harassment for women (Fitzgerald, Shullman, et al., 1988). Many of the items include behaviors that might be enjoyed, especially in informal social settings like those frequently experienced by col- lege students. Therefore, to determine whether an experience was harassing, I asked respondents who experienced an item at least once to indicate how negative (bothersome or stressful) or positive (fun or flattering) it was for them, from –2 (very negative) to �2 (very positive). An experience was considered harassing if the respondent (a) had it at least once and (b) evaluated it negatively. The frequency with which the participant experienced an item (0 to 4) was multiplied by the participant’s evaluation of it, which was set to 0 if the evaluation was neutral or positive (to eliminate from consideration behaviors that had not been experienced negatively), 1 if somewhat negative, and 2 if very negative. The amount of harassment, therefore, ranged from 0 (never experienced or expe- rienced neutrally to positively) to 8 (experienced most of the time very negatively) for each item. Alpha reliability across these scores on the 14 items was .74.

Most investigators have measured harassment by making a priori classifications of certain behaviors as harassing and counting

1 Prior versions of the Sexual Experiences Questionnaire have used different frequency labels. For example, Gelfand, Fitzgerald, and Drasgow (1995) used never, once, and more than once; Fitzgerald, Drasgow, et al. (1997) used a 5-point scale from never to most of the time but did not specify how the middle three values were labeled; Waldo et al. (1998) and Cortina (2001) used never, once or twice, sometimes, often, and most of the time; Fitzgerald et al. (1999) used never, once or twice, sometimes, often, and very often. For comparative purposes, it would be preferable to have standard labels (Gutek, Murphy, & Douma, 2004). The current labels have in common with prior ones the anchors of never to most of the time and the second label of once or twice. The third and fourth labels (a few times and several times) are identical to those used by Berdahl and Moore (2006).

Table 1 Caucasian and Asian College Students’ Ratings of the Desirability of Bem Sex Role Inventory (Short Form) Characteristics for Men and Women in American Society

Item

Caucasian sample (n � 46) Asian sample (n � 85)

Desirability for a man Desirability for a woman Desirability for a man Desirability for a woman

M SD 95% CI M SD 95% CI M SD 95% CI M SD 95% CI

Masculine 5.07 1.11 4.52, 5.62 3.65 0.95 3.28, 4.02 4.94 0.75 4.71, 5.17 4.08 0.92 3.80, 4.36 Feminine 4.03 1.00 3.53, 4.53 5.53 0.82 5.21, 5.84 4.59 1.02 4.27, 4.91 5.41 0.94 5.12, 5.69

428 BERDAHL

how often respondents experienced them. This practice poses a number of problems, including counting behaviors as harassing that were not experienced as such. Berdahl and Moore (2006) were the first to include respondents’ evaluations of their experiences to determine whether they were harassing and to what degree. This approach was used here because it is more consistent with con- ceptualizations of harassment as behavior that threatens and both- ers the recipient (e.g., EEOC, 1980; Einarsen, 2000; Fitzgerald, Swan, & Magley, 1997).

Control variable. Ethnicity served as a control variable. Like being female, belonging to an ethnic minority group is associated with lower status and, therefore, is likely to increase the experience of being harassed (Berdahl & Moore, 2006). A dummy variable was created for ethnic minority by classifying respondents who indicated an Asian, Arab or Middle Eastern, Caribbean, African, Hispanic, or other background (n � 124) as minority (1) and classifying those who indicated a Caucasian background (n � 41) as White (0).2

Results

Descriptive statistics and correlations among the study variables appear in Table 2. Sixty-five percent of the respondents experi- enced at least one episode of sexual harassment in the past 2 years (i.e., negatively experienced at least one of the items at least once). The average amount of harassment, which could range from 0 to 8 and was calculated by multiplying the frequencies by the nega- tive evaluations of events, was .29 (SD � .46). An individual who very negatively experienced 2 of the 14 items one time each, for example, would have a value of .29 on the 0 to 8 scale (calculated [2*2*1]/14). Thirty-five percent of respondents had a value of 0, an additional 58% had a value above 0 but below 1, and increas- ingly smaller proportions had higher values, making the data skewed toward 0. Consistent with prior research, experiencing some harassment was relatively common, but experiencing a lot of it was relatively rare.

Regression models predicting sexual harassment were devel- oped using the GENMOD procedure in SAS and fit using a Poisson distribution (Allison, 1999) because harassment followed a distribution much closer to a Poisson than a normal Gaussian curve. Poisson regressions are suitable for data such as these in which the modal value of the dependent variable is zero, there is only one (positive) tail in its distribution, and the standard devia- tion is greater than the mean (Cameron & Trivedi, 1998).

Regressions on sex (controlling for ethnicity, which was not significant) revealed that women experienced more sexual harass-

ment than men (� � –1.02, p � .0001, d � .51; see Table 3 for percentage harassed, estimated marginal means, and confidence intervals). Women averaged the equivalent of experiencing one item on the survey several times very negatively or six items once or twice somewhat negatively. Men averaged the equivalent of experiencing one item once or twice very negatively or two items once or twice somewhat negatively. Men (M � 1.13, SD � .48) experienced these behaviors more often than women did (M � .83, SD � .59; � � .24, p � .01, d � .56), but on average men evaluated them positively (M � .27, SD � .68), whereas women evaluated them negatively (M � –.25, SD � .71; � � .28, p � .0001, d � .75). Because an experience had to be negatively evaluated to count as harassment, this meant that more women (76%) than men (53%) experienced some amount of sexual ha- rassment.

To test the prediction that women with relatively masculine personalities experience the most sexual harassment, I ran regres- sions on sex– gender dummy variables (masculine woman, androg- ynous woman, feminine woman, undifferentiated woman, mascu- line man, androgynous man, and feminine man, with undifferentiated man as the comparison group)3 and ethnicity. Masculine woman and androgynous woman positively predicted sexual harassment; the other sex– gender dummy variables did not (see Table 4). In other words, being a woman with a relatively masculine personality, regardless of femininity, predicted experi- encing more sexual harassment, consistent with Hypothesis 1. Table 3 shows that high masculine women (women classified as masculine or androgynous) experienced about 1.5 times as much harassment (M � .52) as low masculine women (women classified as feminine or undifferentiated) did (M � .30, d � .48), and three

2 The term minority is technically inaccurate here because a majority of participants were ethnic minorities. As the population in North America becomes more ethnically diverse and Caucasians lose their majority status, this term will become increasingly inappropriate.

3 This was the most direct test of the hypothesis that women with relatively masculine personalities are sexually harassed the most by di- rectly comparing masculine women with the other sex– gender groupings. The median classification approach to masculinity and femininity is con- sistent with the BSRI scoring system and my theoretical rationale. Another way to test the hypothesis would be to run regressions on sex and on masculinity and femininity as continuous variables and to include the two- and three-way interactions between the three variables. I did not predict main effects for masculinity and femininity or an interaction between them, however (there were none), and this is a less consistent and more indirect way of testing the hypothesis.

Table 2 Study 1: Descriptive Statistics and Correlations Among the Study Variables (N � 175)

Variable M SD 1 2 3 4 5

1. Ethnic minority .75 .43 — 2. Woman .66 .50 .19* — 3. Masculinity 4.49 .78 �.35*** �.15* .82 4. Femininity 5.01 .82 .04 .22** .13 .86 5. Sexual harassment .29 .46 �.08 .25*** .07 .08 .74

Note. Reliabilities are shown along the diagonal. * p � .05. ** p � .01. *** p � .001.

429SEXUAL HARASSMENT

times as much harassment as men did (M � .17, d � .77). High masculine women experienced the equivalent of two of the survey items quite negatively several times, low masculine women expe- rienced the equivalent of two items somewhat negatively a few times, and men experienced the equivalent of two items somewhat negatively once or twice.

Discussion

The women in this study experienced more sexual harassment than the men, not because they experienced sexual behaviors more often but because they evaluated these behaviors more negatively, consistent with prior research (e.g., Berdahl et al., 1996; Gutek, 1985). This points to the importance of measuring how a respon- dent evaluates an experience before determining whether it was harassment. Women with relatively masculine personalities, re- gardless of how feminine their personalities also were, experienced significantly more sexual harassment than others, supporting Hy- pothesis 1. Follow-up analyses revealed that this was not because masculine women were more likely than others to evaluate their experiences negatively.4

It is possible, however, that women with relatively masculine personalities are more sensitive to potentially sexually harassing behaviors and, therefore, are more likely to recall them. This is similar to the whiner hypothesis (so termed by Magley, Hulin, Fitzgerald, & DeNardo, 1999), which posits that surveys overes- timate harassment because respondents “whine” about minor events. Research has refuted this hypothesis (e.g., Ilies, Hauser- man, Schwochau, & Stibal, 2003; Magley et al., 1999; Munson, Miner, & Hulin, 2001), but it is likely to be put forth as an explanation for why women with relatively masculine personali- ties are more likely than others to report sexual harassment. It is possible that relatively masculine women experience even fewer potentially harassing behaviors than others but are more likely to recall them when asked. Women who defy gender roles may be more likely to question, be bothered by, and therefore remember behavior based on these roles, such as sexually objectifying women. This provides an alternative hypothesis for why relatively

masculine women were found to experience the most sexual ha- rassment in Study 1:

Hypothesis 2: Women with relatively masculine personalities are more likely than others to negatively evaluate the same potentially harassing behaviors.

A second study was conducted to test this hypothesis. In Study 2, I held the behaviors evaluated by individuals constant by asking them to imagine experiencing the same behaviors measured in Study 1 and to indicate whether they thought they would experi- ence these behaviors negatively, neutrally, or positively. Rather than recalling past experiences, which may involve biases based on sensitivity to these experiences, respondents evaluated identical hypothetical experiences. If relatively masculine women are more negatively sensitive to potentially sexually harassing behaviors, they should evaluate these experiences more negatively.

Study 2

Sample

Participants were 134 students (55 men) enrolled in undergrad- uate management courses at a large public university in the North- east who participated in exchange for course credit. Participants came to a large classroom in groups of approximately 20 and individually completed questionnaires. Seventy-four percent iden- tified their ethnic background as Asian; 14% as Caucasian; 1.5% as Caribbean; and 1% or less as African, Hispanic, Middle Eastern, or other. Average age was 20.77 years (SD � 2.22).

Measures

Personality gender. As in Study 1, personality gender was measured with the short form of the BSRI (Bem, 1978). Individ- uals’ self-ratings on the masculine items were averaged to compute masculinity, and their self-ratings on the feminine items were averaged to compute femininity. Individuals were classified as masculine, androgynous, feminine, or undifferentiated on the basis of their scores relative to within-sex medians, as in Study 1 (male medians: masculinity � 4.80, femininity � 5.10; female medians: masculinity � 4.30, femininity � 5.30; see Table 4 for the number of respondents in each category). Alpha reliabilities for the mas- culine (� � .81) and feminine (� � .81) items were good.

Evaluations of potentially sexually harassing behaviors. The same items used in Study 1 to measure potentially sexually ha- rassing behaviors were used in Study 2. Rather than asking re- spondents to indicate how often they had each experience during the past 2 years, however, I asked respondents to imagine that they had just graduated from college and started a new job and to evaluate how negatively (bothersome or stressful) or positively (fun or flattering) they think they would experience each behavior on that job, from –2 (very negative) to �2 (very positive; � � .91).

4 This measure of harassment represents the product of the frequency and the negative evaluation of potentially harassing behaviors. Follow-up analyses revealed that women with relatively masculine personalities did not experience these behaviors more often than others or evaluate them more negatively but that, when these were combined into a measure of harassment, relatively masculine women were harassed more.

Table 3 Study 1: Experience of Sexual Harassment by Sex and Personality Masculinity

Sex and personality masculinity Harassed (%) Ma SD 90% CIb

Women 76 .41 .45 .32, .48 High masculinec 83 .52 .44 .41, .62 Low masculine 69 .30 .44 .20, .40

Men 53 .17 .44 .09, .25 High masculine 49 .16 .44 .04, .28 Low masculine 57 .18 .44 .07, .29

a Estimated marginal means. b Ninety percent confidence intervals were used because of the one-tailed direction of Hypothesis 1 and the relatively small numbers of observations in each category that resulted from dividing the samples in each study into eight groups on the basis of sex (male and female) and personality gender (masculine, androgynous, feminine, and undifferentiated). c High masculine � masculine (high masculine, low feminine) or androgynous (high masculine, high feminine); low mascu- line � feminine (low masculine, high feminine) or undifferentiated (low masculine, low feminine).

430 BERDAHL

Sexual behaviors are more likely to be seen as wrong and harmful in work than in social settings (cf. Gutek, 1985; Illies et al., 2003). Most students were upper level undergraduates with some work experience who planned to take full-time jobs after graduation. This scenario, therefore, was a more sensitive test of potentially negative reactions to these behaviors and was likely to be mean- ingful and realistic for these participants.

Control variable. As in Study 1, ethnicity was controlled for with a dummy variable for ethnic minority, set to 1 for Asian, African, Caribbean, Hispanic, or Middle Eastern (n � 115) and 0 for Caucasian (n � 19).

Results

Descriptive statistics and correlations among the study variables appear in Table 5. Unlike actual experiences of sexual harassment, which were skewed toward zero, evaluations of hypothetically harassing behaviors followed a normal distribution. Regressions on sex, controlling for ethnicity (which was not significant), re- vealed that women evaluated the behaviors significantly more negatively than men (� � –.58, p � .0001, d � –1.10), although both men and women evaluated them negatively (see Table 6).

To test Hypothesis 2, that women with relatively masculine personalities are more likely than others to negatively evaluate the same potentially harassing behaviors, I ran regressions on sex– gender dummy variables (masculine woman, androgynous woman,

feminine woman, undifferentiated woman, masculine man, an- drogynous man, and feminine man, with undifferentiated man as the comparison group) and ethnicity (see Table 4). The main effect for sex was reflected in the fact that masculine woman, androgy- nous woman, feminine woman, and undifferentiated woman all predicted more negative evaluations, but androgynous man, fem- inine man, and masculine man did not (see Table 6 for estimated marginal means and confidence intervals). Regressions were run on the female data only with undifferentiated woman as the com- parison category and masculine woman, androgynous woman, and feminine woman as predictors (controlling for ethnicity). None was significant. Masculine (� � .01, SE � .18, ns) and androgy- nous (� � .18, SE � .15, ns) women were no more likely than feminine (� � .02, SE � .17, ns) and undifferentiated women to evaluate the potentially harassing behaviors negatively. Therefore, Hypothesis 2 did not receive support.

Discussion

This study demonstrated that women who describe their person- alities in relatively masculine terms are no more likely than others to negatively evaluate potentially sexually harassing events. This suggests that the reason relatively masculine women experienced the most sexual harassment in Study 1 is probably not due to an enhanced likelihood on their part to view these experiences neg-

Table 4 Regressions on Sexual Harassment by Sex and Personality Gender Groups and Control Variables

Variable

Study 1: Actual experiences (N � 175)

Study 2: Hypothetical evaluations (N � 134)

Study 3: Actual experiences (N � 238)

Male-dominated organization (n � 88)

Female-dominated organization (n � 150)

n B SE n B SE n B SE n B SE

Ethnic minority 122 �.41 .26 115 �.08 .15 43 .56 .44 98 .14 .53 Income — — — — 87 �.25 .45 147 �.02 .30 Hours per week — — — — 88 �.31 .50 147 �.04 .33 Masculine woman 19 1.17* .52 15 �.73*** .20 5 2.36*** .71 25 �.10 1.23 Androgynous woman 28 1.40** .49 22 �.55** .18 6 .51 1.02 37 .86 1.04 Undifferentiated woman 32 .99 .51 29 �.73*** .17 7 1.41 .76 44 .15 .80 Feminine woman 19 .77 .57 13 �.70*** .20 5 1.52 .87 27 .52 1.10 Masculine man 18 .01 .66 12 �.15 .23 8 �.07 1.35 3 �21.42 1.00 Androgynous man 17 �.14 .70 15 �.28 .19 19 .24 .81 3 1.30 1.33 Feminine man 20 .51 .58 9 .03 .20 13 .14 .96 5 .59 1.86

* p � .05. ** p � .01. *** p � .001.

Table 5 Study 2: Descriptive Statistics and Correlations Among the Study Variables (N � 134)

Variable M SD 1 2 3 4 5

1. Ethnic minority .86 .35 — 2. Woman .59 .49 .31*** — 3. Masculinity 4.49 .81 �.13 .30*** .81 4. Femininity 5.21 .75 .26** .18* .16 .81 5. Evaluations of potential harassment �1.08 .62 �.19* .48*** .19* �.09 .91

Note. Reliabilities are shown along the diagonal. * p � .05. ** p � .01. *** p � .001.

431SEXUAL HARASSMENT

atively but instead reflects an actual likelihood for them to expe- rience more sexual harassment.

As in Study 1, women evaluated the sexual behaviors more negatively than did men. Whereas men in Study 1 evaluated these behaviors positively, men in Study 2 evaluated them negatively. Women also evaluated the behaviors more negatively in Study 2 than in Study 1. These results are consistent with prior research showing that these behaviors are viewed more negatively in work than in social settings (Gutek, 1985; Illies et al., 2003): Respon- dents in Study 1 were asked to report their experiences in a variety of life domains, whereas respondents in Study 2 were asked to imagine these experiences at work.

Studies 1 and 2 used student samples. Although sexual harass- ment occurs in a variety of social contexts, including school, most research and policy are concerned with sexual harassment at work. Study 3, therefore, was conducted to test whether the results of Study 1 replicate in a work setting, with relatively masculine women experiencing the most sexual harassment on the job.

Study 3

Sample

Surveys were mailed to approximately 800 employees at their home addresses from their union. Each employee worked at one of five organizations located in the same major North American city. Three of the organizations were male-dominated manufacturing plants owned by the same parent company, and two of the orga- nizations were female-dominated community service centers over- seen by the city government. The survey was accompanied by a letter from the union explaining the study, guaranteeing partici- pants’ anonymity and encouraging recipients to complete the ap- proximately 45-min long survey and return it in a postage-paid envelope to the researcher. Participants were paid $15 for com- pleting the survey.

Of the initial mailing, 238 employees completed the survey. This is a fairly typical response rate (30%) for survey research of this nature (e.g., Schneider, Swan, & Fitzgerald, 1997) and is

rather good for a survey of this length and content (cf. Fitzgerald, Weitzman, Gold, & Ormerod, 1988). Of the respondents, 88 (23 women) were employed at a manufacturing plant, and 150 (15 men) were employed at a community center. Modal income was $20,000 to $30,000 per year, modal age was 40 to 49 years, and modal tenure in the organization was 10 to 19 years. Forty-eight percent of the respondents’ ethnic backgrounds were European; 28% were Asian; 10% were Caribbean; 5% were African; 5% were Latin, Central, or South American; and 4% or less were Aborigi- nal, Arab, or Pacific Islander.

Measures

Occupational gender. Dominant sex in the organization was synonymous with dominant sex in the occupation and served as a measure for occupational gender. Employees were represented by the same union and held similar jobs in their organizations. Em- ployees in male-dominated manufacturing plants performed tradi- tionally male-dominated manufacturing jobs, and employees in the female-dominated community centers performed traditionally female-dominated counseling and social work jobs. There were no secretaries surveyed from the male-dominated organizations, for example, or janitors surveyed from the female-dominated ones. This allowed Hypothesis 1 to be tested with a second form of gender conformity and deviance in addition to personality gender by examining whether women in male-typed jobs were sexually harassed more than others.

Personality gender. As in the prior two studies, I measured personality gender with the short form of the BSRI (Bem, 1978). Respondents’ self-ratings on the masculine items (� � .68) were averaged to compute masculinity, and their self-ratings on the feminine items (� � .89) were averaged to compute femininity. Medians for masculinity and femininity were similar in the male- dominated organizations and differed somewhat from those in the female-dominated organizations. Median masculinity was higher and median femininity was lower in male-dominated (masculin- ity � 4.70, femininity � 5.50) than in female-dominated (mascu- linity � 4.60, femininity � 5.60) organizations. Therefore, per- sonality gender differed somewhat by context, consistent with prior research (Prentice & Miller, 2002). To take this into account, as well as the fact that assessments of masculinity and femininity tend to be made within sex (e.g., Connell, 1987; Locksley & Colton, 1979), I calculated medians for masculinity and femininity for men in male-dominated organizations (4.70 and 5.30, respec- tively), women in male-dominated organizations (5.00 and 5.70), men in female-dominated organizations (4.60 and 5.80), and women in female-dominated organizations (4.50 and 5.60). Indi- viduals were classified as masculine, androgynous, feminine, or undifferentiated on the basis of their scores relative to within- organization and within-sex medians (see Table 4 for the number of respondents in each category).

Sexual harassment. Sexual harassment was measured with the same items used in Studies 1 and 2. Respondents indicated how often they experienced each item during the past 2 years at work, from 0 (never) to 4 (most of the time). As in Study 1, if they experienced an item at least once, they were also asked to rate how negative (bothersome or stressful) or positive (fun or flattering) it was for them on a scale from –2 (very negative) to �2 (very positive). Harassment scores were computed for each item by

Table 6 Study 2: Evaluations of Potentially Sexually Harassing Behaviors by Sex and Personality Masculinity

Variable Negative

(%) Ma SD 90% CIb

Women 99 �1.32 .54 �1.42, �1.22 High masculinec 97 �1.26 .55 �1.41, �1.11 Low masculine 100 �1.37 .53 �1.50, �1.24

Men 87 �.74 .55 �.86, �.60 High masculine 87 �.86 .57 �1.04, �.68 Low masculine 87 �.62 .51 �.78, �.46

a Estimated marginal means. b Ninety percent confidence intervals were used because of the one-tailed direction of Hypothesis 2 and the relatively small numbers of observations in each category that resulted from dividing the samples in each study into eight groups on the basis of sex (male and female) and personality gender (masculine, androgynous, feminine, and undifferentiated). c High masculine � masculine (high masculine, low feminine) or androgynous (high masculine, high feminine); low mascu- line � feminine (low masculine, high feminine) or undifferentiated (low masculine, low feminine).

432 BERDAHL

multiplying its frequency by its negative evaluation (converted to absolute values), and these scores were averaged across items. Scale reliability was high (� � .86).

Control variables. Ethnicity, income, and hours worked per week served as control variables. A dummy variable for ethnic minority was created so that respondents who indicated an Asian, Caribbean, African, Latin American, Aboriginal, Arab, or Pacific Islander ethnic origin (n � 123) were labeled minority (1) and those who indicated a European origin (n � 115) were labeled White (0). Annual income was used as a proxy for organizational status, which should affect the likelihood of being harassed. In- come was measured categorically from 1 to 6, with 1 being the lowest income category ($6,500 or less) and 6 being the highest ($45,000 or more). Hours worked per week were controlled for because, all else being equal, the chances of being harassed at work are likely to increase with time spent at work. Hours per week were measured categorically, with 6% of employees report- ing that they worked � 10 hr per week, 6% reporting 11–20 hr, 33% reporting 12–35 hr, 50% reporting 36 – 45 hr, and 4% report- ing � 46 hr per week.

Results

Descriptive statistics and correlations among the study variables appear in Table 7. Twenty-seven percent of the respondents expe- rienced at least one episode of sexual harassment (i.e., negatively experienced at least one item at least once in the past 2 years at work). The average amount of harassment was .14 (SD � .38). As in Study 1, most respondents had a value of 0 and increasingly smaller proportions had higher values, making the data skewed toward 0 and following a Poisson rather than a normal distribution. Regression models predicting the amount of harassment, therefore, were fit using a Poisson distribution (Allison, 1999).

Regressions by sex, occupational gender, and their interaction (with ethnicity, income, and hours worked per week as controls) revealed a significant interaction between sex and occupational gender (� � .26, p � .05). Women in male-typed occupations (M � .50) experienced more sexual harassment than women in female-typed ones (M � .09, d � .90) and more than men in male-typed (M � .06, d � 1.01) and female-typed (M � .08, d � .94) occupations (see Table 8). In other words, women who worked in a male-dominated occupation and organization were more likely than their male counterparts and men and women in female-dominated occupations and organizations to be sexually harassed, consistent with Hypothesis 1. There were no main effects for sex (� � –.02, ns), occupational gender (� � –.03, ns), or the

control variables. As in Study 1, men (M � .47, SD � .51) experienced sexual behaviors more often than women (M � .20, SD � .32; � � –.75, p � .001, d � –.63), but men evaluated them positively (M � .25, SD � .68), whereas women evaluated them negatively (M � –.50, SD � .76; � � –.66, p � .0001, d � –1.04).

To see whether women with relatively masculine personalities were more likely than others to be sexually harassed, I ran regres- sions separately for male-dominated and female-dominated orga- nizations on sex– gender dummy variables (masculine woman, androgynous woman, feminine woman, undifferentiated woman, androgynous man, feminine man, and masculine man, with undif- ferentiated man as the comparison category) and the control vari- ables (ethnicity, income, and hours worked per week).5 In male- dominated organizations, masculine woman, but not the other sex– gender dummy variables, was significant (see Table 4 for regression results and Table 8 for estimated marginal means and confidence intervals). Masculine women (M � .69) averaged more than twice as much harassment as other women (M � .32, d � .54) in male-dominated organizations and more than 8 times as much as men (M � .06, d � .98). Masculine women experienced the equivalent of more than three behaviors several times negatively, whereas other women experienced just over two behaviors a few times negatively and men experienced less than one behavior once or twice negatively. None of the control variables (ethnicity, income, and hours worked per week) was significant. In female- dominated organizations, none of the sex– gender dummy or con- trol variables predicted sexual harassment.

Discussion

Study 3 largely replicated the results of Study 1, but in a work context. As in Study 1, men experienced more sexual behaviors than women, but men tended to enjoy these experiences and

5 Although hierarchical linear modeling may seem appropriate, with organization (or male-dominated vs. female-dominated organization) serv- ing as the higher level variable, at least two properties of these data make hierarchical linear modeling inappropriate (see Hofmann, Griffin, & Gavin, 2000). First, the data are not normally distributed, making traditional maximum-likelihood estimation used by hierarchical linear models inap- propriate. Second, to have sufficient power to detect cross-level interac- tions, a sample of 30 groups (or organizations) or more, with 30 individuals or more each, is recommended. These data include 5 organizations, 2 of which have fewer than 30 respondents. The data, therefore, were analyzed with Poisson regression models for male-dominated and female-dominated organizations separately.

Table 7 Study 3: Descriptive Statistics and Correlations Among the Study Variables (N � 238)

Variable M SD 1 2 3 4 5 6

1. Male-dominated organization .38 .49 — 2. Ethnic minority .59 .49 �.15* — 3. Woman .67 .47 �.64*** .11 — 4. Masculinity 4.60 .94 .07 �.01 �.01 .68 5. Femininity 5.52 .86 .10 �.01 .10 �.30*** .86 6. Sexual harassment .14 .38 .03 .03 .10 .06 .08 .86

Note. Reliabilities are shown along the diagonal. * p � .05. *** p � .001.

433SEXUAL HARASSMENT

women did not. Women who worked in male-dominated jobs at male-dominated manufacturing plants were more likely to experi- ence sexual harassment than their male counterparts in those jobs and men and women in female-dominated jobs at female- dominated community service centers. Women in these male- dominated contexts with relatively masculine personalities expe- rienced the most sexual harassment. In other words, the more women deviated from traditional gender roles— by occupying a “man’s” job or having a “masculine” personality—the more they were targeted for sexual harassment. Although having a masculine personality would seem to help employees fit into male-dominated work environments, having such a personality appears to have hurt the women in this study.

The amount of sexual harassment observed in Study 3 was lower than in Study 1. In Study 1, 65% of respondents experienced at least one episode of sexual harassment and the average amount of harassment was .29 (on a scale of 0 to 8). In Study 3, these numbers were 27% and .14, respectively. This difference is likely due to the fact that these behaviors are viewed as more inappro- priate and, therefore, may be less likely in work than in social settings.

Unlike Study 1 and prior research, there was no main effect for sex on harassment in Study 3. Women were sexually harassed more than men only in male-dominated organizations; within those organizations, only women with masculine personalities were ha- rassed more than others. Had this sample included only employees in male-dominated organizations, like most prior research on sex- ual harassment, there probably would have been a main effect for sex.

General Discussion

These studies provide the first systematic evidence that women who violate feminine ideals are most likely to be sexually harassed in their social and working lives. Study 1 showed that young adult women with relatively masculine personalities (whether or not their personalities also contained feminine characteristics) were more likely than others to be sexually harassed at school, among friends, or at work. Study 3 showed that women who violated gender roles by working in male-dominated occupations experi-

enced more sexual harassment than did men in those occupations and men and women in female-dominated ones. Women who violated feminine ideals for personality accounted for this differ- ence: Masculine women were significantly more harassed in male- dominated domains; feminine women were not significantly more likely than men in these domains to be sexually harassed.

Other research has shown that women in male-dominated job contexts are more likely than others to be sexually harassed (Fitzgerald, Drasgow, et al., 1997; Glomb et al., 1999; Gruber, 1998; Mansfield et al., 1991), but this could be the result of a relatively high amount of sexual attention and interest from men in these environments. The current research suggests that sexual harassment as traditionally defined for women—as consisting of sexual and sexist comments, unwanted sexual attention, and sexual coercion—is primarily targeted at women who step out of place by having masculine characteristics, or “uppity” women. By implica- tion, this suggests that sexual harassment is driven not out of desire for women who meet feminine ideals but out of a desire to punish those who violate them.

If sexual harassment is primarily targeted at women who violate gender roles, it is clear why it is a form of sex discrimination. Viewing sexual harassment as the insensitive or nefarious pursuit of sexual expression and gratification has generated much confu- sion and controversy about why and whether sexual harassment constitutes sex discrimination and has led to policies that focus on policing sexual behavior at work rather than acts that perpetuate sex inequality (Berdahl, in press; Schultz, 1998; C. L. Williams, Giuffre, & Dellinger, 1999). If sexual harassment is seen instead as a pervasive mechanism for punishing gender role violators and thereby a way to enforce different standards of behavior in women and men, then it serves as a basic means for creating different terms and conditions of employment on the basis of sex. Rather than ask whether a perpetrator is sexually motivated or whether the behavior is sexual in content to determine whether it is sexual harassment, this perspective suggests instead that we ask whether an individual of one sex would have been harassed for having the same characteristics and behaving in the same way as an individual of the other sex. Furthermore, it suggests that policy and research on sexual harassment should move beyond focusing on sexual

Table 8 Study 3: Experience of Sexual Harassment by Sex, Personality Masculinity, and Dominant Sex in Organization

Variable

Male-dominated organizations Female-dominated organizations

Harassed (%) Ma SD 90% CIb Harassed (%) Ma SD

90% CIb

Women 57 .50 .39 .37, .63 23 .09 .51 .02, .16 Masculinec 100 .69 .32 .45, .93 21 .04 .40 0, .17 Other 44 .32 .32 .20, .44 24 .14 .40 .08, .20

Men 26 .06 .48 0, .17 20 .08 .50 0, .25 Masculine 25 .05 .31 0, .28 0 .00 .40 0, .38 Other 26 .08 .32 0, .12 25 .16 .40 0, .35

Total 34 .28 .47 .20, .36 23 .09 .82 0, .20

a Estimated marginal means. bNinety percent confidence intervals were used because of the one-tailed direction of Hypothesis 1 and the relatively small numbers of observations in each category that resulted from dividing the samples in each study into eight groups on the basis of sex (male and female) and personality gender (masculine, androgynous, feminine, and undifferentiated). cMasculine individuals were above the median in personality masculinity and below the median in personality femininity within sex.

434 BERDAHL

motives and behaviors to include other forms of behavior that derogate individuals on the basis of sex (Berdahl, in press; Schultz, 1998).

This research suggests that acting like “one of the boys” by being assertive and leader-like may not be the best strategy for women who wish to succeed in male-dominated occupations. Study 1 showed that regardless of how warm or feminine a woman was, being dominant or otherwise masculine was associated with experiencing more harass- ment. Study 3 showed that women who had masculine personality styles but were low on warmth or femininity fared the worst. These results highlight the double bind faced by women who are dismissed and disrespected if feminine but scorned and disliked if masculine, limiting their ascent up the organizational ladder (e.g., Eagly & Karau, 2002; Heilman, 2001). There appears to be little that women can do to avoid being victims of sex discrimination. The onus should not be on victims to avoid a wrong but on those in charge to create structures and incentives to prevent it. A better solution to preventing sexual harassment is to focus on systemic means of discouraging such bias. Employers should focus on eliminating different treatment, standards, and status between male and female employees (Schultz, 2003). Being outspoken or having a traditionally male job should not be accompanied by punishments for women but not for men, for exam- ple. Organizational policies should focus not on banning sexual be- havior per se (Schultz, 1998) but on creating respectful work envi- ronments that do not derogate individuals on the basis of sex and treat men and women with the same characteristics in the same way.

Limitations and Directions for Future Research

There are limitations to the current studies that could be im- proved or complemented by future research. The three studies were based on self-reports of personality gender and sexual ha- rassment experiences. They, therefore, provide correlational but not causal evidence. It is possible, for example, that women who experience the most sexual harassment behave assertively and leader-like because of these experiences. Theory and experimental evidence suggest that this is not the case (e.g., Maass et al., 2003; Rudman, 1998). To rule this out, however, more experimental or longitudinal designs are needed that include other measures of masculinity, femininity, and harassment.

It would have been preferable to have larger sample sizes to study whether masculine women are more likely than others to be targeted for sexual harassment. Such data on sexual harassment are very difficult to obtain in a field setting; therefore, although not ideal, these data provide rare and valuable information. However, once the samples in these studies were divided into eight groups by sex (male and female) and personality gender (masculine, femi- nine, androgynous, or undifferentiated), the number of observa- tions in each cell was relatively small. In Study 3, for example, there were only five to seven women in each gender group in male-dominated organizations. Being a masculine woman signif- icantly predicted sexual harassment, suggesting a strong effect, but conclusions based on these data must be viewed as preliminary because of the small number of women on which they are based. The means of sexual harassment for masculine women and for others differed significantly with 90%, but not with 95%, confi- dence intervals. These differences are likely to have been more significant with larger samples.

This research provides only indirect evidence of what motivates sexual harassment by studying who is targeted and inferring the motives from this. Future research should systematically study what motivates sexual harassment more directly. This has been done by some experimental researchers (e.g., Maass et al., 2003; Pryor, 1987) but is difficult to accomplish with field research. Future field studies could try to examine the motives of harassers by studying their characteristics and the antecedents to their be- havior.

The student and employee samples in these studies all came from one major metropolitan area in the Northeast and were ethnically diverse. Studies 1 and 2 had a majority of Asian par- ticipants, whereas Study 3 had a plurality of Caucasian ones. Future research should examine whether these results generalize to other populations and cultures. Norms for masculinity and femi- ninity vary somewhat across contexts and cultures; therefore, what qualifies as a gender role violation should vary as well.

Estimates of the prevalence of sexual harassment were relatively high in these studies. In Study 1, 76% of the women and 53% of the men had experienced sexual harassment in the past 2 years at school, work, or among family and friends. Cortina et al. (1998) reported that 49% of the women students they surveyed had been sexually harassed in the past 2 years at school. Study 3 showed that 57% of women and 26% of men in male-dominated organizations and 23% of women and 20% of men in female-dominated ones had been sexually harassed in the past 2 years at work. The U.S. Merit System Protection Board study (1995) reported that 44% of women and 19% of men had experienced sexual harassment in the past 2 years at work. The relatively high prevalence rates obtained in the current studies might be due to the fact that these respon- dents reported on their experiences with same-sex as well as other-sex perpetrators.

Conclusion

This research provides the first systematic evidence based on field data that sexual harassment is primarily targeted at women who violate gender ideals. This highlights the role of sexual harassment as a form of sex discrimination that keeps the sexes separate and unequal at work (Schultz, 1998). Women may behave in “feminine” ways and avoid behaving in “masculine” ones because they face negative re- percussions when they do not (e.g., Eagly, Makhijani, & Klonsky, 1992; Rudman, 1998). More research is needed to document and understand the means by which gender roles get enforced to perpet- uate difference and inequality between the sexes. Future research should explore whether forms of sexual harassment identified against men are mainly targeted at men who violate gender ideals (Berdahl et al., 1996; Franke, 1997; Stockdale et al., 1999). Unlike harassment in general (Brodsky, 1976; Einarsen, 2000), sex-based harassment has been conceptualized as an active but not a passive harm (Schultz, 1998). Future research on sex-based harassment should include the study of passive forms it may take, such as social exclusion, ignoring, and slander.

References

Allison, P. D. (1999). Logistic regression using the SAS system: Theory and application. Cary, NC: SAS Institute.

Alvesson, M., & Billing, Y. D. (1997). Understanding gender and orga- nizations. London, England: Sage.

435SEXUAL HARASSMENT

American Heritage Dictionary of the English Language. (4th ed.). (2004). Boston: Houghton Mifflin.

Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42, 155–162.

Bem, S. L. (1978). The Bem Sex Role Inventory. Palo Alto, CA: Consulting Psychologists Press.

Berdahl, J. L. (in press). Harassment based on sex: Protecting social status in the context of gender hierarchy. Academy of Management Review.

Berdahl, J. L., Magley, V. J., & Waldo, C. R. (1996). The sexual harass- ment of men: Exploring the concept with theory and data. Psychology of Women Quarterly, 20, 527–547.

Berdahl, J. L., & Moore, C. (2006). Workplace harassment: Double jeop- ardy for minority women. Journal of Applied Psychology, 91, 426 – 436.

Bergen, D. J., & Williams, J. E. (1991). Sex stereotypes in the United States revisited: 1972–1988. Sex Roles, 24, 413– 423.

Brodsky, C. M. (1976). The harassed worker. Toronto, Canada: Lexington Books.

Burgess, D., & Borgida, E. (1999). Who women are, who women should be: Descriptive and prescriptive gender stereotyping in sex discrimina- tion. Psychology, Public Policy, and the Law, 5, 665– 692.

Buss, D. M. (1989). Sex differences in human mate preferences: Evolu- tionary hypotheses tested in 37 cultures. Behavioral and Brain Sciences, 12, 1– 49.

Cameron, A. C., & Trivedi, P. K. (1998). Regression analysis of count data. Cambridge, England: Cambridge University Press.

Campbell, T., Gillaspy, J. A., Jr., & Thompson, B. (1997). The factor structure of the Bem Sex-Role Inventory (BSRI): Confirmatory analysis of long and short forms. Educational and Psychological Measurement, 57, 118 –124.

Connell, R. W. (1987). Gender and power. Stanford, CA: Stanford Uni- versity Press.

Connell, R. W. (1995). Masculinities. Berkeley: University of California Press.

Cortina, L. M. (2001). Assessing sexual harassment among Latinas: De- velopment of an instrument. Cultural Diversity and Ethnic Minority Psychology, 7, 164 –181.

Cortina, L. M., Swan, S., Fitzgerald, L. F., & Waldo, C. (1998). Sexual harassment and assault: Chilling the climate for women in academia. Psychology of Women Quarterly, 22, 419 – 441.

Dall’Ara, E., & Maass, A. (1999). Studying sexual harassment in the laboratory: Are egalitarian women at higher risk? Sex Roles, 41, 681– 704.

Duffy, J., Wareham, S., & Walsh, M. (2004). Psychological consequences for high school students of having been sexually harassed. Sex Roles, 50, 811– 821.

Eagly, A. H. (1987). Sex differences in social behavior: A social-role analysis. Hillsdale, NJ: Erlbaum.

Eagly, A. H., & Karau, S. J. (2002). Role congruity theory of prejudice toward female leaders. Psychological Review, 109, 573–598.

Eagly, A. H., Makhijani, M. G., & Klonsky, B. G. (1992). Gender and the evaluation of leaders: A meta-analysis. Psychological Bulletin, 111, 3–22.

Eagly, A. H., Wood, W., & Diekman, A. B. (2000). Social role theory of sex differences and similarities: A current appraisal. In T. Eckes & H. M. Trautner (Eds.), The developmental social psychology of gender (pp. 123–174). Mahwah, NJ: Erlbaum.

Einarsen, S. (2000). Harassment and bullying at work: A review of the Scandinavian approach. Aggression and Violent Behavior, 5, 379 – 401.

Equal Employment Opportunity Commission. (1980). Guidelines on dis- crimination because of sex (Sect. 1604.11). Federal Register, 45, 74676 –74677.

Evans, L. J. (1978). Sexual harassment: Women’s hidden occupational hazard. In J. Chapman & M. Gates (Eds.), The victimization of women (pp. 203–223). Beverly Hills, CA: Sage.

Farley, L. (1978). Sexual shakedown: The sexual harassment of women on the job. New York: McGraw-Hill.

Fiske, S. T., Cuddy, A. J. C., Glick, P., & Xu, J. (2002). A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82, 878 –902.

Fiske, S. T., & Glick, P. (1995). Ambivalence and stereotypes cause sexual harassment: A theory with implications for organizational change. Jour- nal of Social Issues, 51, 97–115.

Fiske, S. T., & Stevens, L. E. (1993). What’s so special about sex? Gender stereotyping and discrimination. In S. Oskamp & M. Costanzo (Eds.), Gender issues in contemporary society (pp. 173–196). Newbury Park, CA: Sage.

Fitzgerald, L. F., Drasgow, F., Hulin, C. L., Gelfand, M. J., & Magley, V. J. (1997). Antecedents and consequences of sexual harassment in organi- zations: A test of an integrated model. Journal of Applied Psychology, 82, 578 –589.

Fitzgerald, L. F., Gelfand, M. J., & Drasgow, F. (1995). Measuring sexual harassment: Theoretical and psychometric advances. Basic and Applied Social Psychology, 17, 425– 445.

Fitzgerald, L. F., Magley, V. J., Drasgow, F., & Waldo, C. R. (1999). Measuring sexual harassment in the military: The Sexual Experiences Questionnaire (SEQ–DoD). Military Psychology Special Issue: Sexual Harassment, 11, 243–263.

Fitzgerald, L. F., Shullman, S. L., Bailey, N., Richards, M., Swecker, J., Gold, A., et al. (1988). The incidence and dimensions of sexual harass- ment in academia and the workplace. Journal of Vocational Behavior, 32, 152–175.

Fitzgerald, L. F., Swan, S., & Magley, V. J. (1997). But was it really sexual harassment? Legal, behavioral, and psychological definitions of the workplace victimization of women. In W. O’Donohue (Ed.), Sexual harassment: Theory, research, and treatment (pp. 5–28). Needham Heights, MA: Allyn & Bacon.

Fitzgerald, L. F., Weitzman, L. M., Gold, Y., & Ormerod, M. (1988). Academic harassment: Sex and denial in scholarly garb. Psychology of Women Quarterly, 12, 329 –340.

Franke, K. M. (1997). What’s wrong with sexual harassment? Stanford Law Review, 49, 691–772.

Gelfand, M. J., Fitzgerald, L. F., & Drasgow, F. (1995). The structure of sexual harassment: A confirmatory analysis across cultures and settings. Journal of Vocational Behavior, 47, 164 –177.

Glomb, T. M., Munson, L. J., Hulin, C. L., Bergman, M. E., & Drasgow, F. (1999). Structural equation models of sexual harassment: Longitudi- nal explorations and cross-sectional generalizations. Journal of Applied Psychology, 84, 14 –28.

Gruber, J. E. (1998). The impact of male work environments and organi- zational policies on women’s experiences of sexual harassment. Gender and Society, 12, 301–320.

Gutek, B. A. (1985). Sex and the workplace: The impact of sexual behavior and harassment on women, men, and organizations. San Francisco: Jossey-Bass.

Gutek, B. A., Cohen, A. G., & Konrad, A. M. (1990). Predicting social– sexual behavior at work: A contact hypothesis. Academy of Management Journal, 33, 560 –577.

Gutek, B. A., & Morasch, B. (1982). Sex ratios, sex-role spillover, and sexual harassment of women at work. Journal of Social Issues, 38, 55–74.

Gutek, B. A., Murphy, R. O., & Douma, B. (2004). A review and critique of the Sexual Experiences Questionnaire (SEQ). Law and Human Be- havior, 28, 457– 482.

Harris v. Forklift Sys. (92–1168), 510 U.S. 17 (1993). Heilman, M. E. (2001). Description and prescription: How gender stereo-

types prevent women’s ascent up the organizational ladder. Journal of Social Issues, 57, 657– 674.

Hofmann, D. A., Griffin, M. A., & Gavin, M. B. (2000). The application

436 BERDAHL

of hierarchical linear modeling to organizational research. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions (pp. 467– 511). San Francisco, CA: Jossey-Bass.

Holt, C. L., & Ellis, J. B. (1998). Assessing the current validity of the Bem Sex-Role Inventory. Sex Roles, 39, 929 –941.

Ilies, R., Hauserman, N., Schwochau, S., & Stibal, J. (2003). Reported incidence rates of work-related sexual harassment in the United States: Using meta-analysis to explain reported rate disparities, Personnel Psy- chology, 56, 607– 631.

Koss, M. P., Goodman, L. A., Browne, A., Fitzgerald, L. F., Keita, G. P., & Russo, N. F. (1994). No safe haven: Male violence against women at home, at work, and in the community. Washington, DC: American Psychological Association.

Lengnick-Hall, M. L. (1995). Sexual harassment research: A methodolog- ical critique. Personnel Psychology, 48, 841– 864.

Locksley, A., & Colton, M. E. (1979). Psychological androgyny: A case of mistaken identity? Journal of Personality and Social Psychology, 37, 1017–1031.

Maass, A., Cadinu, M., Guarnieri, G., & Grasselli, A. (2003). Sexual harassment under social identity threat: The computer harassment par- adigm. Journal of Personality and Social Psychology, 85, 853– 870.

MacKinnon, C. A. (1979). Sexual harassment of working women. New Haven, CT: Yale University Press.

Magley, V. J., Hulin, C. L., Fitzgerald, L. F., & DeNardo, M. (1999). Outcomes of self-labeling sexual harassment. Journal of Applied Psy- chology, 84, 390 – 402.

Mansfield, P. K., Koch, P. B., Henderson, J., Vicary, J. R., Cohn, M., & Young, E. W. (1991). The job climate for women in traditionally male blue-collar occupations. Sex Roles, 25, 63–79.

McCormack, A. (1985). The sexual harassment of students by teachers: The case of students in science. Sex Roles, 13, 21–32.

Meritor Savings Bank v. Vinson (84 –1979) 477 U.S. 57 (1986). Minkowitz, D. (1994, April 19). Love hurts: Brandon Teena was a woman

who lived and loved as a man. She was killed for carrying it off. The Village Voice, p. 24.

Munson, L. J., Miner, A. G., & Hulin, C. (2001). Labeling sexual harass- ment in the military: An extension and replication. Journal of Applied Psychology, 86, 293–303.

Nieva, V. F., & Gutek, B. A. (1981). Women and work: A psychological perspective. New York: Praeger.

O’Leary-Kelly, A. M., Paetzold, R. L., & Griffin, R. W. (2000). Sexual harassment as aggressive behavior: An actor-based perspective. Acad- emy of Management Review, 25, 372–388.

Paludi, M. A. (1997). Sexual harassment in schools. In W. O’Donohue (Ed.), Sexual harassment: Theory, research, and treatment (pp. 225– 240). Needham Heights, MA: Allyn & Bacon.

Prentice, D., & Miller, D. T. (2002). The emergence of homegrown stereotypes. American Psychologist, 57, 352–359.

Prentice, D. A., & Carranza, E. (2002). What women and men should be, shouldn’t be, are allowed to be, and don’t have to be: The contents of prescriptive gender stereotypes. Psychology of Women Quarterly, 26, 269 – 281.

Price Waterhouse v. Hopkins, 109 S. Ct. 1775 (1989). Pryor, J. B. (1987). Sexual harassment proclivities in men. Sex Roles, 17,

269 –290. Pryor, J. B., LaVite, C., & Stoller, L. (1993). A social psychological

analysis of sexual harassment: The person/situation interaction. Journal of Vocational Behavior, 42, 68 – 83.

Reilly, M. E., Lott, B., & Gallogly, S. M. (1986). Sexual harassment of university students. Sex Roles, 15, 333–358.

Rudman, L. A. (1998). Self-promotion as a risk factor for women: The costs and benefits of counterstereotypical impression management. Jour- nal of Personality and Social Psychology, 74, 629 – 645.

Rudman, L. A., & Glick, P. (1999). Feminized management and backlash toward agentic women: The hidden costs to women of a kinder, gentler image of middle managers. Journal of Personality and Social Psychol- ogy, 77, 1004 –1010.

Rudman, L. A., & Glick, P. (2001). Prescriptive gender stereotypes and backlash toward agentic women. Journal of Social Issues, 57, 743–762.

Sanchez v. City of Miami Beach, 720 F. Supp. 974 (S.D. Fla. 1989). Schacht, S. P., & Atchison, P. H. (1993). Heterosexual instrumentalism:

Past and future directions. Feminism & Psychology, 3, 37–53. Schneider, K. T., Swan, S., & Fitzgerald, L. F. (1997). Job-related and

psychological effects of sexual harassment in the workplace: Empirical evidence from two organizations. Journal of Applied Psychology, 82, 401– 415.

Schultz, V. (1998). Reconceptualizing sexual harassment. Yale Law Jour- nal, 107, 1683–1796.

Schultz, V. (2003). The sanitized workplace. Yale Law Journal, 112, 2061–2193.

Stockdale, M. S., Visio, M., & Batra, L. (1999). The sexual harassment of men—Evidence for a broader theory of sexual harassment and sex discrimination. Psychology, Public Policy, and the Law, 5, 630 – 664.

Studd, M. V., & Gattiker, U. E. (1991). The evolutionary psychology of sexual harassment in organizations. Ethology and Sociobiology, 12, 249 –290.

Tangri, S., & Hayes, S. M. (1997). Theories of sexual harassment. In W. O’Donohue (Ed.), Sexual harassment: Theory, research, and treatment (pp. 99 –111). Needham Heights, MA: Allyn & Bacon.

U.S. Merit Systems Protection Board. (1981). Sexual harassment in the federal workplace: Is it a problem? Washington DC: U.S. Government Printing Office.

U.S. Merit Systems Protection Board. (1988). Sexual harassment in the federal government: An update. Washington DC: U.S. Government Printing Office.

U.S. Merit Systems Protection Board. (1995). Sexual harassment in the federal workplace: Trends, progress, continuing challenges. Washing- ton, DC: U.S. Government Printing Office.

Waldo, C. R., Berdahl, J. L., & Fitzgerald, L. F. (1998). Are men sexually harassed? If so, by whom? Law and Human Behavior, 22, 59 –79.

Welsh, S. (1999). Gender and sexual harassment. Annual Review of Soci- ology, 25, 169 –190.

Williams, C. L., Giuffre, P. A., & Dellinger, K. (1999). Sexuality in the workplace: Organizational control, sexual harassment, and the pursuit of pleasure. Annual Review of Sociology, 25, 73–93.

Williams, J. E., & Best, D. L. (1990). Sex and psyche: Gender and self viewed cross-culturally. Newbury Park, CA: Sage.

Williams v. Bell, 587 F. 2d 1240 (D.C. Cir. 1978). Zalk, S. R. (1990). Men in the academy: A psychological profile of

harassment. In M. Paludi (Ed.), Ivory power: Sexual harassment on campus (pp. 141–175). Albany: State University of New York Press.

Received June 24, 2005 Revision received April 13, 2006

Accepted July 10, 2006 �

437SEXUAL HARASSMENT