Archives of Sexual Behavior

, Volume 41, Issue 3, pp 691–702

Gender and Binegativity: Men’s and Women’s Attitudes Toward Male and Female Bisexuals


    • Departments of Psychology and Women’s & Gender StudiesDickinson College
    • Department of PsychologyDickinson College
  • Genéa D. Thomas
    • Department of PsychologyDickinson College
Original Paper

DOI: 10.1007/s10508-011-9767-8

Cite this article as:
Yost, M.R. & Thomas, G.D. Arch Sex Behav (2012) 41: 691. doi:10.1007/s10508-011-9767-8


This study assessed the influence of gender on attitudes about bisexuals. A total of 164 heterosexual female and 89 heterosexual male undergraduates completed the Biphobia Scale (Mulick & Wright, 2002), rewritten to refer to bisexual men and bisexual women and thus re-named the Gender-Specific Binegativity Scale. A mixed-design ANOVA revealed an interaction between rater’s sex and target’s sex: women equally accepted bisexual men and bisexual women, but men were less accepting of bisexual men than bisexual women. A mediation analysis indicated the relationship between rater’s sex and greater acceptance of bisexual women was partially explained by eroticization of female same-sex sexuality. Finally, participants also responded to two open-ended items, which provided additional information about the content of binegativity: participants described male bisexuals negatively, as gender-nonconforming, and labeled them “really gay,” whereas participants described female bisexuals positively, as sexy, and labeled them “really heterosexual.” These findings suggest multiple underlying beliefs about bisexuals that contribute to binegativity, particularly against bisexual men. Results also confirm the importance of considering gender (of both the target and the rater) when assessing sexual prejudice.


BisexualityBiphobiaSexual prejudiceEroticism of lesbiansGender differences


Although prejudiced attitudes against gay men and lesbians have been well-documented in the United States over the past several decades (Herek, 1998), prejudiced attitudes against bisexuals have only recently been examined (Eliason, 1997, 2001; Mohr & Rochlen, 1999; Mulick & Wright, 2002). This lack of attention to attitudes toward bisexuals is unfortunate, particularly given recent documentation that bisexuality is more common than lesbianism in women in the U.S. (Laumann, Gagnon, Michael, & Michaels, 1994; Mosher, Chandra, & Jones, 2005), and that bisexual sexual attractions are more common than exclusively homosexual sexual attractions in young men and women (Savin-Williams & Ream, 2007).

The purpose of the present study was to contribute to the literature on sexual prejudice against bisexuals, by exploring the influence of gender on these attitudes. In developing hypotheses, we drew upon the few studies that have examined attitudes toward bisexuals, and we drew heavily upon research comparing men’s and women’s sexual prejudice against lesbians and gay men.

Biphobia and Binegativity

Biphobia has been defined as a set of prejudiced attitudes about individuals with a bisexual sexual orientation. The term was first coined by Bennett (1992), who stated that biphobia represented “the denigration of bisexuality as a valid life choice” (p. 207). However, many scholars have moved away from the use of the term “homophobia” in favor of more accurate terms such as “sexual prejudice” (Herek, 2003) or “homonegativity” (Hudson & Ricketts, 1980) when discussing attitudes about gay men and lesbians. This shift in terminology reflects the understanding that such prejudiced attitudes do not, in fact, involve a phobia or fear. Rather, these attitudes toward gay and lesbian individuals are negative cognitions and affective reactions, which also often include lack of support for civil rights for gay men and lesbians. To be consistent with current conceptualizations, for the purpose of the present research, we will use the term “binegativity,” not biphobia, when discussing prejudiced attitudes about bisexuals.

Scholars and activists have theorized about the content of binegativity, noting that it is expressed by homosexuals and heterosexuals alike (Weiss, 2004). Binegativity includes the belief that bisexuals are confused about their sexuality, are immature, or are “really” gay or lesbian; expressions of doubt that bisexuality exists; concern that bisexuals (specifically, bisexual men) are responsible for spreading HIV/AIDS to the heterosexual community; and the belief that bisexuals need the companionship of both male and female partners in order to feel sexually or emotionally complete, thereby implying that bisexuals are sexually promiscuous and unable to be monogamous (Farajajé-Jones, 1995; Ochs, 1996; Rust, 2002; Sumpter, 1991).

Although the activist literature on binegativity is fairly extensive, little empirical research has been conducted. A few studies have assessed binegativity while specifying gender, but most relied on only a single-item assessment, which is not psychometrically desirable. For example, Herek (2002) conducted a “feeling thermometer” survey with a representative sample of adults in the U.S. and found that participants felt more positively about bisexual women than they did about bisexual men. Eliason (1997) similarly found more positive attitudes about bisexual women than men among heterosexual undergraduates in the U.S., using a single item of “tolerance.” Spalding and Peplau (1997) provided participants with vignettes describing dating couples (with a male bisexual target in a relationship with a gay man or a heterosexual woman, and a female bisexual target in a relationship with a lesbian or a heterosexual man). They found that participants judged bisexuals as less monogamous and more likely to give a sexually transmitted infection to a partner; this study did not find differences regarding the bisexual person’s gender.

Rating scales are psychometrically preferable to single-item measures. The Attitudes Regarding Bisexuality Scale (ARBS; Mohr & Rochlen, 1999) contains subscales referring to male and female bisexuals separately, and measures two aspects of binegativity–tolerance (viewing bisexuality as an acceptable sexual orientation) and stability (viewing bisexuality as a stable sexual orientation and not a phase). However, the ARBS was limited in that these subscales only represent cognitions (beliefs and stereotypes about a group), whereas Eagly and Chaiken (1993) have shown that prejudices are multifactorial, involving cognitions as well as affect and behaviors. Thus, we believe it is important to explore gender differences in binegativity using a measure that reflects the multifactorial nature of prejudice.

The Biphobia Scale (Mulick & Wright, 2002) is more comprehensive, including items reflecting negative cognitions, negative affect, and discriminatory behaviors (both behavioral avoidance and acting out), and thus represents an improvement over the ARBS. But the Biphobia Scale itself is limited in that the items specify attitudes toward “a bisexual individual.” Since studies of gender-neutral language indicate that when gender is unspecified, participants assume maleness (Hamilton, 1991; Merritt & Kok, 1995), it is probable that the Biphobia Scale has inadvertently assessed attitudes toward bisexual men. The Biphobia Scale was chosen for use in the present study due to its greater range of items, but modification of the scale (to specify gender of the bisexual) was undertaken in order to explore our research questions.

Gender and Homonegativity

The importance of studying sexual prejudice with attention to gender can be easily seen by turning to the extensive literature on sexual prejudice against lesbians and gay men. It is well-documented that women are more accepting than men of sexual minorities (Britton, 1990; Logan, 1996; Reiter, 1991). Various scholars of men and masculinity have even argued that heterosexual masculinity is characterized by homonegativity (Herek, 1986; Kimmel, 1994), providing a theoretical explanation for men’s greater sexual prejudice than women’s.

In addition, a number of scholars argue that the best research on sexual prejudice assesses attitudes toward lesbians or gay men separately, because the target’s sex has an important influence on rater’s attitudes (Herek, 1994; Kite & Whitley, 2003). Although there are a few scales of homonegativity that are gender neutral (Hudson & Ricketts, 1980; Wright, Adams, & Bernat, 1999), most recent scales of homophobia and heterosexism, such as Attitudes about Lesbians and Gay Men (Herek, 1994) and the Modern Homophobia Scale (Raja & Stokes, 1998), specify “lesbian” or “gay man” in the items. Thus, these scales assess prejudice based on sexual orientation while simultaneously assessing the impact of target’s sex.

Evidence of the importance of considering lesbians and gay men separately comes from meta-analytic research that finds that heterosexual participants hold more positive attitudes about lesbians than gay men (Kite & Whitley, 1996). This may be due to the greater sanctions men face when violating expectations around masculinity (Stockard & Johnson, 1979); because homosexuality is seen as a gender violation, gay men are judged more harshly than lesbian women. In addition, more positive attitudes about lesbians compared to gay men may be due to the eroticization of lesbians by many heterosexual men (Reiss, 1986). Since heterosexual women do not eroticize gay male sexuality to the same degree that heterosexual men eroticize lesbian sexuality (Nyberg & Alston, 1977), the positivity associated with men’s sexualization of female same-sex sexuality attenuates what would otherwise be more negative attitudes toward lesbians. In fact, Louderback and Whitley (1997) found that undergraduate men’s eroticization of lesbians explained their greater acceptance of lesbians than gay men, as shown by the fact that when the researchers controlled for erotic value of lesbian sex, men’s ratings of lesbians and gay men became roughly equivalent.

The Present Study

The primary purpose of the present study was to assess whether the pattern of attitudes previously found for lesbians and gay men were similar when considering bisexual women and men. We undertook the present research by modifying the Biphobia Scale (Mulick & Wright, 2002) to specify the bisexual target’s sex, which we re-named the Gender-Specific Binegativity Scale (GSBS). This allowed us to explore three interrelated hypotheses about gender and binegativity: (1) Participants would express greater binegativity against male than female bisexuals; (2) Male participants would express greater binegativity than female participants; and, (3) Male participants would express greater binegativity against male bisexuals than against female bisexuals, whereas female participants’ binegativity would not differ based on the target’s sex.

The second purpose of the present research was to explore a possible reason for gender differences in sexual prejudice. We wondered whether eroticization of lesbian sexuality might play a role in more positive attitudes toward bisexual women, as this has been shown to be a reason for men’s more positive attitudes toward lesbians (Louderback & Whitley, 1997). Our fourth hypothesis stated that eroticization of lesbian sexuality would mediate the relationship between rater’s sex and differential attitudes toward bisexual men and women.

Lastly, we included open-ended items asking participants about bisexual men and women. These items were designed to provide additional information about gender differences in binegativity. Our fifth hypothesis stated that open-ended items would reveal additional beliefs about bisexuals that will also vary depending on both rater’s and target’s sex.



Participants were 164 heterosexual women and 89 heterosexual men from a small, liberal arts college in Pennsylvania. Ages ranged from 18 to 25 years (M = 19.46, SD = 1.36). The majority of participants were White (79.1%, 127 women and 73 men); 14 (5.5%) participants were biracial or multiracial, 13 (5.1%) were African American, 9 (3.6%) were Latino/a, 7 (2.8%) were Asian American, and 10 participants declined to provide their ethnic background.1

This study was IRB approved and utilized the department’s research participant pool for recruitment, with students receiving credit for their participation. Advertising materials indicated that volunteers were needed for a study on “college students’ beliefs about sexuality and sexual relationships,” but no mention was made of our focus on bisexuality. Participants completed the questionnaire in small groups (20 participants or fewer). Participants were allotted 30 minutes to complete the questionnaire, including the open-ended items.


Open-ended Items

Participants were first asked to respond to two open-ended items: “When you think of a bisexual woman, what comes to mind?” and “When you think of a bisexual man, what comes to mind?” These items were presented in the same order for all participants.

We coded the responses thematically, using an exploratory, grounded approach in which we initially read through all the responses to develop a list of in vivo codes utilizing the participants’ own words (following the guidelines for thematic analysis as delineated by Braun & Clarke, 2006). This initial step was conducted by the two authors collaboratively. We then created a coding scheme using the codes generated in the initial reading; each coding category was scored as present (1) or absent (0) in each participant’s responses.

Defining Bisexuality

We coded “defining bisexuality” if a participant explained that a bisexual was someone who was sexually or romantically attracted to both men and women. Other definitions for bisexuality (such as explaining that a bisexual was someone who did not use gender as a defining characteristic for their sexuality; see Diamond, 2008b) were also coded.

Really Homosexual

We coded for “really homosexual” whenever a participant indicated that a bisexual person truly has a homosexual (gay, lesbian) sexual orientation.

Really Heterosexual

We coded for “really heterosexual” whenever a participant indicated that a bisexual person truly has a heterosexual (“straight”) sexual orientation.

Wanting Attention

We coded for “wanting attention” when participants indicated that bisexuals engage in bisexual (or same-sex) sexuality in order to get attention from another person. Phrases such as “showing off” and “attention seeker” were coded in this category.


We coded for “masculine” when a participant described a female bisexual with the term “masculine” or mentioned other indications of traditional masculinity, such as “short hair,” or masculine social identities, such as “tomboy” or “butch.”


We coded for “feminine” when a participant described a male bisexual with the term “feminine” or mentioned other indications of traditional femininity, such as “long hair,” or feminine social identities, such as “femme” or “metrosexual.”


We coded “liberal” when a participant described bisexuals as liberal and open to new experiences.


We coded “sexy” when a participant indicated that bisexuality was sexy, erotic, or “hot.”


We coded “negative” when a participant indicated that bisexuality was wrong, immoral, or disgusting.

Rating Procedure

The first author coded all of the responses and the second author coded 50% of the responses. The observed proportion of overall agreement was extremely high (overall agreement = .98; proportion of agreement ranged from .93 to 1.00 for individual codes), and interrater reliability was very good (kappa = .82, p < .001; kappa ranged from .65 to 1.00 for individual codes). Given the very strong correspondence between our coding and the relative straightforwardness of our coding scheme, we determined that double-coding the entire data set was unnecessary. The few disagreements that occurred for this half of the data set were resolved through discussion, and slight modifications to the coding scheme were made in order to ensure consistent coding for the second half of participants.

To ease interpretation of the findings, we grouped the nine in vivo codes into four overarching themes that reflected conceptual overlap among the codes. Defining Bisexuality remained a single theme and included only the coding category of “defining bisexuality.” Second, the theme Doubting Bisexuality was constructed; we considered that the coding categories “really homosexual,” “really heterosexual,” and “wanting attention” all indicated disbelief that bisexuality exists. The first two codes indicate the belief that a bisexual person was hiding their “true” orientation while the third suggests that the bisexual person was not motivated by their own internal desires, but rather engaged in same-sex sexuality to gain attention from an observer. Third, the code of “masculinity” in reference to a female bisexual and the code of “femininity” in reference to a male bisexual were grouped under the theme of Gender Nonconformity. Finally, we considered the coding categories of “liberal” and “sexy” to reference positive characteristics and “negative” to reference negative characteristics, so we grouped these three codes into the theme Positive and Negative Attributes.

The Gender-Specific Binegativity Scale

Following the open-ended items, participants responded to a revised version of Mulick and Wright’s (2002) Biphobia Scale: the Gender-Specific Binegativity Scale. The original Biphobia Scale contained 30 items assessing negative attitudes and cognitions about bisexuals, as well as prejudiced behaviors toward bisexuals. In the present study, the 28 items that mentioned a bisexual person were modified in order to specify the target’s sex. The Biphobia Scale contains two items that refer to “bisexuality” as an orientation rather than to a bisexual person, and these items were left as written; for example, “I think bisexuality is wrong.” These two items were presented only once, in the men’s version; however, to be consistent with the original scale scoring, these two items were included in the computation of both the GSBS-Men scale score and the GSBS-Women scale score. Presentation of the two versions was counterbalanced.

Each item was assessed on a 6-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). After reverse coding 13 items that were positively valenced, the 30 items referring to a female bisexual were averaged, and the 30 items referring to a male bisexual were averaged. Higher scores indicated greater binegativity.

Eroticization of Lesbian Sexuality

Participants were asked to indicate their arousal in response to sexually explicit visual material (specified to include magazines, videos, and websites). After first indicating how often one viewed such material, participants were presented with the following two questions: “If you do not view pornography, how arousing do you think the following images would be?” and “If you do view pornographic material, how arousing do you find pornography with the following themes to be?” Seven different pornographic themes were presented, with the theme of interest being, “A woman having sex with another woman.” Response options ranged from 1 (“not at all arousing”) to 5 (“extremely arousing”).

A single variable assessing the Eroticization of Lesbian Sexuality was created by separating participants by their level of viewing pornography. If the participant never or rarely viewed pornography, then the response to first item was used; if the participant sometimes or regularly viewed pornography, then the response to the second item was used. Higher scores indicated a greater eroticization of women having sex with one another.


Reliability and Validity of the GSBS

Prior to conducting substantive analyses, we considered order effects on the GSBS-Men via a 2 (Rater’s Sex: Male, Female) × 2 (Order: GSBS-Men first, GSBS-Women first) ANOVA. This analysis showed that attitudes about bisexual men were more negative when participants were presented the GSBS-Men first rather than second (M = 1.84, SD = .78; M = 1.70, SD = .72, respectively), F(1, 249) = 5.44, p < .05. For attitudes toward bisexual women, a second 2 (Rater’s Sex) × 2 (Order) ANOVA on the GSBS-Women showed that attitudes about bisexual women did not significantly differ depending on order (M = 1.70, SD = .70 for men presented first; M = 1.61, SD = .58 for women presented first), F(1, 249) = 3.22, p = .07.

Next, we examined the psychometric properties of the GSBS-Men and the GSBS-Women. To determine scale reliability, we computed alpha coefficients for each of these two new scales. Our results were in line with the alphas reported by Mulick and Wright (2002) on the original measure: coefficient alpha was .95 for the GSBS-Men and .93 for the GSBS-Women.

Finally, to demonstrate that the two new scales retained the validity of the original scale despite our modifications to specify the target’s gender, we conducted two separate factor analyses (GSBS-Men, GSBS-Women). Factor analysis of the original Biphobia Scale (Mulick & Wright, 2002) had yielded a one-factor solution.

The factor structure of the GSBS-Men was examined via exploratory factor analysis (EFA) using Maximum Likelihood Solution (Harman, 1976). An oblique rotation of the matrix was undertaken to allow the factors to correlate with one another rather than forcing an orthogonal solution (see Kline, 1994).

A total of 238 participants responded to all items, which is a ratio of 7.9 participants per item. The common factors were overdetermined and items showed an average communality of .60, which was indicative of sufficient participants for a valid EFA (MacCallum, Widaman, Zhang, & Hong, 1999). Further evidence for scale factorability was demonstrated using Bartlett’s test of sphericity (χ2 = 4597.70, p < .001), and the Kaiser–Meyer–Olkin measure (KMO = .93, which exceeds the minimum of .60 recommended by Tabachnick & Fidell, 2001). These indicators justified the use of EFA with this sample.

We initially identified five factors with eigenvalues greater than 1, which together explained 62% of the variance. However, one factor accounted for 43% of the variance, and no other factor accounted for more than 6.4% of the variance; the scree plot dropped off sharply after the first factor. Gorsuch (1983) argued that a one-factor model is appropriate when at least 20% of the variance is accounted for by the first factor, which our analysis met. The factor analysis was re-run specifying a one-factor solution. All factor loadings were at or above .38 (items with their loadings are displayed in Table 1).
Table 1

Factor loadings for the Gender-Specific Binegativity Scale


Factor loadings

Women items

Men items


I do not like bisexual women/men




I think bisexuality is wrong




I would like to have a bisexual woman/man as a neighbor




I would be friends with a woman/man who is bisexual




I am comfortable around bisexual women/men




I discriminate against bisexual women/men




I would hit a bisexual woman/man for coming on to me




Bisexual women/men spread AIDS to the heterosexual population




Bisexual women/men make me nervous




Bisexual women/men deserve to get discriminated against




Bisexuality is acceptable to me




I do not think that bisexual women/men should work with children




I make derogatory remarks about bisexual women/men




Bisexual women/men should not get married




Bisexual women/men are not capable of monogamous relationships




I would be comfortable having a bisexual woman/man as a roommate




I tease and joke about bisexual women/men




You cannot trust a woman/man who is bisexual




I would get angry if a bisexual woman/man made sexual advances toward me




I think I could work with a bisexual woman/man




I get anxious when I have to interact with bisexual women/men




I avoid bisexual women/men




When I meet a bisexual woman/man, I think, “what a waste”




I have rocky relationships with women/men I suspect are bisexual




Bisexual women/men want to have sex with everybody




Bisexual women/men are not capable of controlling their sexual impulses




I feel uneasy around bisexual women/men




I would not go to a public place where I knew there would be bisexual women/men




It does not matter to me if my female friends are bisexual




I would not want to talk to a woman/man who I knew was bisexual



Note: Items 3, 4, 5, 11, 16, 20, and 29 were reverse-scored prior to computing scale scores

Next, the factor structure of the GSBS-Women was explored with the same analysis. A total of 243 participants responded to all items; this factor analysis had a ratio of 8.10 participants per item, overdetermined factors, and average communalities of .58. Further evidence for scale factorability was demonstrated using Bartlett’s test of sphericity (χ2 = 4075.93, p < .001) and KMO (.88). These indicators justified the use of EFA with this sample.

We identified seven factors with eigenvalues greater than 1, which together explained 66% of the variance. One factor accounted for 36% of the variance, and no other factor accounted for more than 8.3% of the variance; the scree plot dropped off sharply after the first factor. Because Gorsuch’s criterion for a one-factor solution was met, we re-ran the factor analysis specifying a one-factor solution. All items loaded on the single factor with values at or above .37, with the exception of the item, “I would hit a bisexual woman for coming on to me” (factor loading = .26) (see Table 1).

These analyses indicated that the GSBS-Men and GSBS-Women retained the psychometric properties that the original Biphobia scale demonstrated (Mulick & Wright, 2002), with evidence of a one-factor solution (for each target sex) that combined all attitudinal items.

Descriptive Statistics

Means and SD for all scales are shown in Table 2. Overall, participants rejected binegativity. We conducted four single-sample t-tests to compare the average score of male and female raters on both the GSBS-Men and the GSBS-Women with the neutral midpoint of 3.5. Male raters scored significantly lower than the neutral midpoint on the GSBS-Men, t(88) = 14.66, p < .001, and on the GSBS-Women, t(88) = 22.82, p < .001. Female raters also scored significantly lower than the neutral midpoint on the GSBS-Men, t(163) = 41.50, p < .001, and on the GSBS-Women, t(163) = 40.89, p < .001.
Table 2

Descriptive statistics for all scales, by rater’s sex


Rater’s sex



n = 89

n = 164















Eroticization of lesbian sexb





aAbsolute range, 1–6; b Absolute range, 1–5

Purpose 1: Gender and Binegativity

We hypothesized that: (1) attitudes about bisexual women would be more positive (less prejudiced) than attitudes about bisexual men, (2) male participants would express greater binegativity than would female participants, and (3) male participants will be more prejudiced against male bisexuals in comparison with female bisexuals. A 2 (Target’s Sex) × 2 (Rater’s Sex) ANOVA was conducted.

In support of Hypothesis 1, the ANOVA showed a significant main effect of Target’s Sex: participants were more positive toward bisexual women than bisexual men, F(1, 251) = 57.84, p < .001 (large effect size, η2 = .19, observed power 1.00).

In support of Hypothesis 2, this analysis also showed a significant main effect of Rater’s Sex: women were more positive than were men toward bisexuals, F(1, 251) = 21.43, p < .001 (medium effect size, η2 = .08, observed power .99).

Finally, in support of Hypothesis 3, there was a significant interaction between Rater’s Sex and Target’s Sex, F(1, 251) = 66.81, p < .001 (large effect size, η2 = .21, observed power 1.00).2 This interaction is displayed in Fig. 1.
Fig. 1

Interaction between rater’s sex and the bisexual person’s gender

To probe the interaction, we conducted four simple comparisons of means. Female participants were equally positive toward bisexual women and men, paired t(163) < 1, but male participants were more positive toward bisexual women than men, paired t(88) = 8.14, p < .001, large effect size (Cohen’s d = .86). In addition, independent samples t-tests showed that male and female participants differed somewhat in their binegativity against bisexual women, t(251) = 2.46, p < .05, small effect size (Cohen’s d = .31), but they differed more strongly in their binegativity against bisexual men, t(251) = 6.31, p < .001, large effect size (Cohen’s d = .79).

Purpose 2: Understanding the Gender Differences

Hypothesis 4: Eroticization of Lesbian Sex

The ANOVA indicated that, as expected, men reported more positive attitudes about bisexual women than about bisexual men, whereas women reported equivalent attitudes about bisexuals regardless of their gender. Our fourth hypothesis was that the eroticization of lesbian sexuality would mediate the relationship between target’s sex and more positive evaluations of bisexual women.

Before conducting the series of multiple regression analyses that would test mediation, the data required recoding. First, gender was dummy coded so that the “0” indicated women and “1” indicated men. Second, for the outcome variable, we created a Difference Score by subtracting the GSBS-Women from the GSBS-Men. This resulted in scores that could range from −5 (greater binegativity against female than male bisexuals) to +5 (greater binegativity against male than female bisexuals). A score at or near 0 would indicate that attitudes toward bisexual men and women were approximately equal.

We followed Baron and Kenny’s (1986) guidelines to test mediation using regression (see Fig. 2). In Step 1, the simple regression analysis with Rater’s Sex predicting the Difference Score was significant (R2 = .21, p < .001). Rater’s Sex significantly predicted the Difference Score, with men more likely than women to express greater binegativity against male than female bisexuals.
Fig. 2

Four-step mediation analysis, showing standardized regression coefficients

In Step 2, the simple regression analysis with Rater’s Sex predicting the Eroticization of Lesbian Sex was significant (R2 = .28, p < .001). Rater’s Sex significantly predicted the Eroticization of Lesbian Sex, showing that men eroticized lesbian sex more than women. An examination of the means bears this out: an independent samples t-test showed that men’s scores on Eroticization of Lesbian Sex were significantly higher than women’s scores, t(251) = 9.91, p < .001, large effect size (Cohen’s d = 1.33) (see Table 2).

In Step 3, the simple regression analysis with the Eroticization of Lesbian Sex predicting GSBS Difference was significant (R2 = .17, p < .001). Eroticization of Lesbian Sex significantly predicted the GSBS Difference Score, with greater Eroticization of Lesbian Sex associated with larger Difference Scores (greater binegativity against male than female bisexuals).

In Step 4, the multiple regression with Rater’s Sex and Eroticization of Lesbian Sex predicting the Difference Score was significant (R2 = .25, p < .001), and collinearity statistics were within acceptable limits (tolerance = .72, VIF = 1.39, condition indices < 6). Rater’s Sex significantly predicted the GSBS Difference Score, and greater Eroticization of Lesbian Sex also significantly predicted the GSBS Difference Score. Here, the mediator predicted the outcome variable even when accounting for the effect of Rater’s Sex; thus, the mediation was supported.

However, the relationship between Rater’s Sex and the Difference Score was reduced but not eliminated when we accounted for Eroticization of Lesbian Sex, which indicated partial mediation. Rater’s Sex significantly predicted the GSBS Difference score on its own (β = .46 in Step 1), and when accounting for the mediator (β = .33 in Step 4). Thus, men’s (as compared to women’s) greater acceptance of female than male bisexuality was only partially explained by the fact that men eroticized lesbian sexuality more than women.

Hypothesis 5: Additional Gender Differences in Binegativity

Finally, we explored participants’ open-ended responses to items about bisexual men and women, in the hopes that these responses could provide additional detail about the beliefs held by participants. Theoretically, we assumed that the perceptions about bisexual men and women that participants wrote represented influences on their self-reported attitudes on the GSBS, and, at the very least, demonstrated additional aspects of binegativity that were not captured quantitatively.

A total of 249 (98%) participants responded to both open-ended items. The mean word length per participant in response to bisexual women was 19.65 (SD = 14.48) and in response to bisexual men was 19.73 (SD = 14.53). A 2 (Target’s Sex) × 2 (Rater’s Sex) ANOVA did not find significant main effects for bisexual target sex, rater sex, or their interaction for word length.

These analyses follow the structure of those conducted with the GSBS. First, we analyzed Target’s Sex: we report chi-square analyses comparing the proportion of participants who scored for a given coding category in reference to bisexual men versus a bisexual women. In order to test whether there was a statistically significant difference, the McNemar statistic was computed. Second, we analyzed Rater’s Sex: we report standard chi-squares comparing the proportion of men and the proportion of women who described a given theme to see whether open-ended responses varied based on the gender of the participant. See Table 3 for the percentages of men and women reporting each theme.
Table 3

Thematic analysis of open-ended items, separated by target’s and rater’s sex


Male bisexual target

Female bisexual target

Male rater

Female rater

Total (all raters)

Male rater

Female rater

Total (all raters)


35 (41%)

101 (62%)

136 (55%)

36 (41%)

108 (66%)

144 (57%)

Really heterosexual

1 (1%)

1 (1%)

2 (1%)

5 (6%)

9 (6%)

14 (6%)

Really gay

18 (21%)

25 (15%)

43 (17%)

3 (4%)

2 (1%)

5 (2%)

Wanting attention

3 (3%)

9 (6%)

12 (5%)




Gender non-conforming

13 (15%)

29 (18%)

42 (19%)

5 (6%)

3 (2%)

8 (3%)

Liberal and open

8 (9%)

13 (8%)

21 (8%)

16 (18%)

25 (15%)

41 (16%)





10 (11%)

6 (4%)

16 (6%)


10 (12%)

5 (3%)

15 (6%)

5 (6%)

2 (1%)

7 (3%)

Note: Number of participants is provided first, followed by percentage in parentheses


The most common response to the open-ended questions about bisexual men and women was definitional: a statement about what it means to be bisexual in terms of attraction or sexual behavior. The McNemar test pairing definition about male bisexuals with definition about female bisexuals showed that participants were equally likely to provide a definition in both cases, McNemar (1, n = 249) = 2.67.

Regarding Rater’s Sex, a definitional response was more frequently provided by women than by men in response to both items: referring to bisexual women, χ2(1, n = 252) = 14.55, p < .001; referring to bisexual men, χ2(1, n = 249) = 10.27, p = .001.

Doubting Bisexuality

Participants were more likely to doubt a bisexual man’s sexual orientation than a bisexual woman’s, McNemar (1, n = 249) = 15.36, p < .001. However, participants were more likely to call a bisexual man “really gay” than a bisexual woman, McNemar (1, n = 249) = 34.38, p < .001, and participants were more likely to call a bisexual woman “really heterosexual” than a bisexual man, McNemar (1, n = 249) = 10.29, p < .01.

The code “wanting attention” was only mentioned in reference to female bisexuals; no participants indicated that a male bisexual wanted attention (McNemar test could not be calculated because “wanting attention” for bisexual men was 0).

Regarding Rater’s Sex, male and female participants did not differ on any of the codes within this theme.

Gender Nonconformity

McNemar’s test pairing gender nonconformity in male bisexuals and in female bisexuals showed that participants were much more likely to judge male bisexuals as gender nonconforming, McNemar (1, n = 249) = 32.11, p < .001. Male and female participants were equally likely to judge bisexuals as gender nonconforming.

Positive and Negative Attributes

Larger numbers of participants attributed positive characteristics (liberal and open to new experiences, and sexy) to female bisexuals than male bisexuals: not one participant said that male bisexuals were sexy, McNemar (1, n = 249) = 16.67, p < .001 for liberal, but incalculable for sexy.

In contrast, a marginally greater proportion of participants attributed negative characteristics to male than to female bisexuals, McNemar (1, n = 249) = 16.67, p = .06.

Regarding Rater’s Sex, men and women were equally likely to judge “liberal and open,” but male participants were more likely than female participants to claim that female bisexuals were sexy, χ2(1, n = 252) = 5.72, p < .05. Male participants were also more likely than female participants to claim that male bisexuality was negative, χ2(1, n = 249) = 7.29, p < .01), and that female bisexuality was negative, χ2 = 4.22, Fisher’s Exact test (used due to small n), p = .05.


Participants generally rejected binegativity as measured by the GSBS; yet, gender differences in mean levels of this form of sexual prejudice still appeared, justifying the argument that gender is highly salient in binegativity. We found that participants were more positive toward bisexual women than men, and women were more positive than men toward bisexuals. Also, whereas female participants were equally accepting of bisexual women and men, male participants more strongly expressed sexual prejudice toward bisexual men than toward women. Additionally, the meditational regression analysis considering participants’ eroticization of women’s same-sex sexuality and the examination of the open-ended responses shed additional light on binegativity.

Target’s Sex

Consistent with previous research on homophobia against gay men and lesbians (Kite & Whitley, 1996), participants in the present study expressed greater binegativity against bisexual men than women. The open-ended responses included some beliefs that can help us understand why bisexual men were judged more harshly than bisexual women.

First, when participants explained that bisexuality was not a “real” sexual orientation, they did so in gender differentiated ways, dismissing the bisexual orientation when describing men but not women. This makes clear that the belief that bisexuality is not a “real” sexual orientation (Ochs, 1996; Sumpter, 1991) was felt more strongly about men. Perhaps one reason why participants expressed greater binegativity against men than women was in reaction to this presumed deception; if participants believed that a bisexual man was lying about his sexual orientation, then part of the binegativity against men might be due to a distaste for dishonesty.

Additionally, responses regarding the probability that male bisexuals were gay indicated that raters believed that any sexual desire or contact with another man was enough to indicate homosexuality. On the other hand, responses about women focused on the probability that female bisexuals were heterosexual. If participants believed bisexual men were really gay but bisexual women were really heterosexual, then binegativity toward men might truly reflect homonegativity, and positive attitudes toward bisexual women might reflect homonegativity’s absence: bisexual women were given the benefit of the doubt and granted some heterosexual privilege.

One belief expressed only about bisexual women was “wanting attention.” Perhaps due to the eroticization of lesbian sexuality, specifically scenes in which sex between women culminates in a threesome with a male partner (Reiss, 1986), bisexual women were thought to be motivated by a desire for male attention. The perception that bisexual women were really heterosexual was then reinforced by the belief that bisexual women were performing for the purposes of attention.

Bisexual men were more likely to be judged as gender non-conforming than were bisexual women. There is ample evidence that some expressions of masculinity in girls and women is increasingly socially acceptable, but that expressions of femininity in boys and men remains stigmatized (Kane, 2006; Sandnabba & Ahlberg, 1999). If participants were less accepting of gender non-conformity in men than in women, then the greater binegativity toward bisexual men could be related to these beliefs about bisexuals’ gender presentations.

A final relevant theme in the open-ended responses was that bisexual women were more likely to be judged as “liberal and open to new experiences” than were men. Assuming that our participants value this type of openness and freedom of expression, and the emerging adulthood literature suggests that college students do value exploration (Arnett, 2004), then our finding may be related to the perception that bisexual women were fun, experimental, free-spirits.

Rater’s Sex

Consistent with previous research on homonegativity, men expressed greater binegativity than did women. As Kite and Whitley (2003) argued, pressures on men to express traditional masculinity, along with the belief that homosexuals transgress traditional gender roles, contribute to men expressing greater homonegativity. Participants in the present sample believed that bisexual men were feminine, which suggests that men’s stronger binegativity may be related to gender role beliefs; men’s binegativity may be driven by distaste for gender non-conformity.

Another explanation relies on different cultural understandings about men’s and women’s sexual orientation. Research by Diamond (2008a) has shown that many women understand their sexuality as fluid, responsive to different social contexts and different partners. Perhaps the sex difference in binegativity was due to the fact that heterosexual women perceive that they might one day be bisexual or that they once were bisexual, whereas heterosexual men perceive that they are, have been, and always will be heterosexual. One may be less likely to be prejudiced against a group that one once belonged to or to which one might belong in the future.

In support of this interpretation, within our heterosexual sample, 5.5% of women had past sexual experiences with women. These 9 women expressed less binegativity against women than did women without same-sex experience (M = 1.30 and 1.60, respectively; t approached statistical significance). Future research could examine whether this difference is reliable with larger samples. It would also be interesting to see whether binegativity relates to certainty about one’s sexual orientation; the Measure of Sexual Identity Exploration and Commitment (Worthington, Navarro, Savoy, & Hampton, 2008), which includes a subscale assessing uncertainty about one’s sexual orientation identity, would be an excellent tool for such an examination.

In contrast to women, men tend to view their sexual orientation as fixed, unchangeable, and set from birth (Otis & Skinner, 2004). Men tend to identify as gay or heterosexual earlier than women (Savin-Williams & Diamond, 2000) and maintain this identity throughout their lifetime whereas women are more likely to shift (Baumeister, 2000). The conception of men’s sexuality as stable and falling into non-overlapping categories (gay or heterosexual) may indicate a degree of discomfort with constructs, like bisexuality, that do not fit neatly into binary categories.

We have some evidence that support the contention that men were less comfortable with the idea of bisexuality. Women more strongly rejected the GSBS item, “I think bisexuality is wrong,” than did men, and women more strongly endorsed the GSBS item, “Bisexuality is acceptable to me,” than did men. Also, in the open-ended responses, men were more likely than women to state that bisexuals were wrong and disgusting. It is already known that women are more tolerant of ambiguity than are men (Rotter & O’Connell, 1982), but it would be interesting to explore whether tolerance for ambiguity would correlate with binegativity but not homonegativity. This would allow researchers to more fully understand a potentially unique aspect of binegativity–negative attitudes about those who live in the “gray areas” between heterosexuality and homosexuality.

Finally, social psychological research on prejudice suggests that men’s greater rates of prejudice are, in fact, due to gender differences in the more general ideological construct of social dominance orientation. This construct refers to the belief that society should be organized according to social hierarchies, and this belief is more strongly endorsed by men than women (Pratto, Sidanius, Stallworth, & Malle, 1994). Men’s greater binegativity may be yet another indicator of underlying social dominance orientation. Future research assessing social dominance alongside binegativity (and with other potentially-relevant predictors of binegativity, such as right-wing authoritarianism, religiosity, and erotophobia) could explore this possibility.

Target’s Sex and Rater’s Sex Interaction

The interaction between the bisexual target’s sex and rater’s sex was significant; this finding was consistent with past research on homonegativity. Herek (2000) and Kite and Whitley (1996) found that heterosexual women reported similar levels of homonegativity toward lesbians and gay men, but men reported stronger homonegativity toward gay men than lesbians. These findings mirror our results on bisexuals: heterosexual women reported similar levels of binegativity toward male and female bisexuals, but men reported stronger binegativity against bisexual men than women.

The mediation analysis indicated that the eroticization of lesbian partially mediated the relationship between rater’s sex and greater binegativity against men: when participants of either sex found lesbian pornography arousing, they were more accepting of bisexual women than men; moreover, men were more likely to eroticize lesbian sexuality than were women on the pornography measure. We similarly found evidence in the open-ended responses that female bisexuality was eroticized whereas male bisexuality was not, and men were more likely than women to state that bisexual women were sexy. This set of findings aligned with Louderback and Whitley’s (1997) analysis on lesbians: although male participants expressed greater acceptance of lesbians than gay men, after controlling for the erotic value of lesbian sex, the difference disappeared. However, it is important to note that, in the present study, the relationship between rater’s sex and binegativity was only partially mediated by the eroticization of lesbian sexuality, suggesting that gender remains an important predictor of stronger binegativity against bisexual men than women.

Limitations and Future Directions

This study was limited due to the relatively homogeneous sample: participants were mostly White college students at a small liberal arts college. Because attitudes about sexuality differ based on demographic variables, such as age, education, and ethnicity, and in different cultural contexts, additional research using the GSBS with more diverse samples is needed.

For instance, in the open-ended responses, no one stated that bisexuals spread HIV/AIDS to heterosexuals, even though activists have contended that this is a common belief (Farajajé-Jones, 1995). Young adults today did not experience the height of the AIDS epidemic, so they may worry less about HIV/AIDS transmission than older adults (see Valdiserri, 2004, on AIDS complacency). The lack of racial diversity in our sample also may have limited responses. Scholars studying African American communities have noted the “down low” phenomenon–heterosexually-identified men who engage in sexual activity with men in secret (Lapinski, Braz, & Maloney, 2010). This might well influence African American’s perceptions of bisexual men, particularly in terms of spreading HIV/AIDS to female partners. Future research with racially diverse samples could ascertain racial differences in attitudes toward bisexual men and women.

Age might also explain the low scores that we found on the GSBS. Young adults tend to report more permissive attitudes about sexuality (Le Gall, Mullet, & Shafighi, 2002) and are more supportive of same-sex marriage (CNN/Opinion Research Corporation, 2010) than older adults, so the findings of low levels of prejudice may be due to our young sample. Future research that includes older adults may result in a wider range of attitudes about bisexuals.

Future research that includes lesbians and gay men could test for differences in mean levels of binegativity based on the rater’s sex and sexual orientation. This would provide a more complete picture of binegativity, considering that bisexual activists have argued that such prejudice comes from both heterosexual and lesbian and gay communities (Tucker, 1995).

Finally, this research relied on self-report of attitudes. Participants were able to discern the intent of the items, and were able to modify their responses to be socially-desirable. Our findings may have under-estimated participants’ actual binegativity. Future research on implicit attitudes about bisexuals, or questionnaire research that controls for social desirability is warranted.


An initial comparison between the White participants and the participants of color on the three attitude scales of interest (GSBS-Men, GSBS-Women, and Eroticization of Lesbian Sex) showed no racial differences (all t values <1.67, all ps > .09). Substantive analyses proceeded without considering race.


Levene’s tests indicated that these data violated the assumption of equality of variances (GSBS-Men p < .001; GSBS-Women p < .05), with the men’s sample showing a larger variance than the women’s sample in both cases. However, after applying an inverse transformation and re-analyzing the data, the results of the ANOVA remained the same (F values dropped slightly, but all p values remained <.001). We concluded that heterogeneity of variance did not unduly affect our results and, therefore, all analyses rely on the untransformed values.


Copyright information

© Springer Science+Business Media, LLC 2011