Gender Nonconformity Is Perceived Differently for Cisgender and Transgender Targets

Abstract

The present research examined the role gender non-conformity plays in attitudes toward transgender people. Study 1 with 232 U.S college students focused on attitudes toward female targets; Study 2 with 217 U.S. college students focused on male targets; and Study 3 with 462 mTurk workers directly compared attitudes toward female and male targets. In all three studies, participants read a vignette depicting either a transgender or cisgender target who presents as either gender conforming or gender nonconforming. In all three studies, we found that gender nonconforming targets and transgender targets were perceived as more threatening to the distinction between men and women, and in two of the studies, we found that gender conforming transgender targets were more threatening than conforming cisgender targets. We also found that anti-transgender prejudice, traditional gender role beliefs (Studies 1 and 2), and biological gender essentialism (Study 3) moderated these effects. Transgender targets who conform to the traditional binary gender role associated with their gender expression are perceived as transgressing distinct binary gender boundaries, which may be threatening because “passing” transgender individuals are harder to detect as transgressors and because their “passing” challenges the belief that gender is biologically essential and immutable. Furthermore, as anti-transgender prejudice, traditional gender role beliefs, and gender essentialist beliefs increase, liking decreased and threat increased for transgender and gender nonconforming targets. Working to alter gender essentialist beliefs may be a route to reducing anti-transgender prejudice.

Prejudice toward and discrimination against transgender individuals (i.e., those whose assigned birth sex is not congruent with their current gender identity; Tate et al. 2014) is widespread. Indeed, fully 58% of transgender individuals report experiences of serious discrimination, including job loss and eviction, bullying and physical harassment, and denial of medical services (James et al. 2016). One source of prejudice against transgender individuals is that they may be perceived as gender nonconforming (i.e., not conforming to the gender roles traditionally assigned to men and women, such as feminine men or masculine women) and thus as threatening toward the gender binary. People evaluate those who do not conform to traditional gender roles more negatively (Blashill and Powlishta 2009; Levy et al. 1995; Richardson et al. 1980), especially nonconforming men (McCreary 1994; Sirin et al. 2004).

However, transgender individuals may be perceived as gender nonconforming in multiple ways—not conforming to the traditional roles of either their assigned sex (i.e., their sex assigned at birth by a medical professional, commonly based on genitalia) or their gender identity (i.e., their authentic gender, which does not conform to their assigned sex). For example, a transgender woman (i.e., assigned as male at birth, but lives as a woman) may be seen as transgressing gender roles for her assigned sex (i.e., male) by identifying and living as a woman, but could also be seen as transgressing gender norms for her gender identity (i.e., woman) if she does not appear or behave femininely. Thus, the question is, would a masculine transgender woman (or feminine transgender man) be perceived the same way as a masculine cisgender (i.e., one whose assigned birth sex is congruent with their current gender identity) woman (or feminine cisgender man)?

Although prior research has examined the connection between violation of the gender binary and anti-transgender prejudice, it has been correlational in nature and has not investigated how gender nonconformity of transgender individuals to their gender identity affects perceptions of them. The aim of our research was to examine attitudes toward cisgender (people whose gender identity and expression matches their assigned birth sex) and transgender (people whose gender identity and expression does not match the label assigned at birth) men and women based on their conformity to gender roles for their gender.

Anti-Transgender Prejudice

Most of the existing anti-transgender prejudice research is correlational and has focused on identifying a number of individual differences that predict anti-transgender prejudice, most often using college or adult samples in the United States. Prior studies have suggested a variety of individual differences and personality traits associated with increased anti-transgender prejudice including right-wing authoritarianism, religious fundamentalism, political conservatism, religiosity (among cisgender women participants only), anti-egalitarian attitudes, and hostile sexism (Nagoshi et al. 2008; Norton and Herek 2013), as well as Need for Closure, preference for traditional gender roles, and anti-gay, anti-lesbian, and anti-bisexual (LGB) prejudice (Makwana et al. 2017; Tebbe and Moradi 2012). Researchers have also found that, overall, men tend to hold more negative attitudes toward transgender individuals than women do (Antoszewski et al. 2007; Norton and Herek 2013; Tebbe and Moradi 2012).

Another potential reason for prejudice and discrimination against transgender individuals is binary gender role socialization because transgender individuals may be seen as threatening the distinction between groups. Gender binarism belief refers to the belief that there are only two genders, corresponding with biological sex (Tebbe and Moradi 2012), and such beliefs are associated with greater anti-transgender prejudice (Ching and Xu 2018; Norton and Herek 2013; Tebbe and Moradi 2012). It is possible that transgender people are threatening to those scoring higher in gender binarism because transgender people are perceived as transgressing the strict boundary between the binary genders.

Social identity theory (Tajfel and Turner 1979) posits that people define themselves based on the social group to which they belong. People tend to see themselves as more similar to other ingroup members and as very different from outgroup members when “us versus them” dynamics are made salient. Similarly, social categorization theory (Turner et al. 1987) states that people determine which social identities to utilize for categorization based on how well the social categories are reflective of actual group differences, preferring categorizations that maximize intragroup similarities and minimize intergroup similarities. Both social identity theory and self-categorization theory (Tajfel and Turner 1979; Turner 1999; Turner et al. 1987) suggest that group members are highly motivated to maintain intergroup distinctiveness (i.e., distinct binary gender boundaries: male versus female) and may attempt to protect such distinctiveness by devaluing outgroup members who are perceived as obscuring group boundaries, a phenomenon known as distinctiveness threat (e.g., Jetten et al. 1996, 2004, 2005; Spears et al. 1997; Van Knippenberg and Ellemers 1990). Transgender individuals may be perceived as transgressing the distinctions between binary gender categories (i.e., men and women) that are based on traditional gender roles (i.e., distinctiveness threat; Glotfelter and Anderson 2017). Thus, individuals who challenge the idea that birth sex solely determines gender identity may threaten binary gender boundaries and challenge gender binarism beliefs. It is possible that transgender people are devalued as a way to protect distinctiveness between the binary genders and uphold gender binarism beliefs (Glotfelter and Anderson 2017).

Additionally, it is possible that the degree to which transgender individuals conform to the traditional gender roles ascribed to their gender expression or identity may affect perceivers’ attitudes toward them. Using data from the 2008 National Transgender Discrimination survey, Miller and Grollman (2015) found that transgender people who reported being perceived as transgender or gender nonconforming (to their gender expression) were more likely to report experiences of major discrimination events and everyday transgender discrimination, even after controlling for medical and social transition status and other demographic factors such as age, race, and gender identity. In effect, transgender people who were identified as transgender based on their appearance (i.e., nonconforming to their gender expression) were more likely to be actively discriminated against. It is possible that transgender individuals who do conform to the traditional gender roles associated with their gender expression are less likely to be actively discriminated against but may be more threatening to binary gender distinctiveness because they are harder to visually distinguish from a cisgender individual.

Gender Conformity

Research on sexism and anti-gay prejudice suggests that people tend to be more prejudiced toward violators of traditional gender roles, especially in appearance and behaviors (Blashill and Powlishta 2012; Rudman 2005). There is ample evidence that gender nonconforming men experience prejudice and backlash in work environments (Moss-Racusin et al. 2010), and that, beyond adult work settings, men may face greater backlash for expressing stereotypically feminine characteristics (e.g., communal, caring, interdependent) than women do for expressing typically male characteristics (e.g., agentic, ambitious, self-reliant; Rudman and Phelan 2008). Indeed, men who transgress behavioral gender roles are often viewed more negatively than are gender role-transgressing women (McCreary 1994).

In addition, gender conformity is often researched in conjunction with sexual orientation. A pervasive (but arguably incorrect) assumption among laypersons and researchers is that gender nonconformity (especially in boys and men) indicates homosexuality (Gottschalk 2003; Herek 1984, 1994). Assumptions of sexual orientation based on gender conformity are strongest for feminine male targets (compared to masculine female targets), especially among heterosexual male raters (McCreary 1994). Furthermore, the stigmatization of gender nonconformity appears to play a role in perceptions of gay people. Feminine gay men are rated as less desirable than masculine gay men, and masculine gay women are rated as less desirable (e.g., acceptable, likeable, “correct”) than feminine gay women, especially by heterosexual male participants (Schope and Eliason 2004). Likewise, gender atypical (i.e., masculine women, feminine men) gay targets were rated as more immoral than gender atypical heterosexual targets, and gay targets were seen as more gender atypical, but only for those raters high in anti-gay prejudice (Lehavot and Lambert 2007). Additionally, people tend to perceive gender atypical behaviors and appearance as not only prescriptive of gay sexual orientation, but also as a conscious demonstration of their gay social identity (Lick et al. 2014). Targets depicting gender atypical walking patterns and facial cues were more likely to be categorized as gay and were presumed to have changed their appearance in order to communicate their sexual orientation to others, which led to more negative evaluations—especially for gay female targets (Lick et al. 2014).

However, some evidence suggests that negative attitudes toward gender conformity and sexual orientation are separate. Blashill and Powlishta (2009) had heterosexual male participants rate gender conforming or gender nonconforming gay men and heterosexual men, and they found that feminine men were rated more negatively (i.e., liked less, avoided more, less desirable as a work partner) and as possessing negative traits (i.e., boring, unintelligent), regardless of sexual orientation. Gay men, regardless of gender conformity, were rated more negatively as well. However, when no information was explicitly given about the target’s sexual orientation, more feminine male targets were rated as more homosexual (although not as “clearly gay”) than were more masculine male targets.

Social identity theory suggests that the socially accepted categories of male/masculine and female/feminine are conflated (i.e., gender identification and gender-role characteristics are based on the same categorization criteria). Thus, an individual who can be categorized as a man or a woman based on their self-identification is expected to also possess gender-conforming characteristics (i.e., femininity for a woman, masculinity for a man). Those who transgress binary gender category expectations by identifying as one gender but having characteristics of the other (e.g., masculine women, feminine men) are likely seen as threatening to binary gender category distinctions and are discriminated against as a means of maintaining distinct binary gender boundaries (i.e., distinctiveness threat).

Gender conformity in transgender targets is potentially more complex because the same behavior or characteristics may be seen as conforming to their gender identification and expression, but nonconforming to their assigned birth sex (or vice versa). One experimental study found that participants rated targets whose facial appearance was incongruent with their gender expression (e.g., a transgender woman who looked more masculine; a transgender man who looked more feminine) more negatively than targets whose facial appearance was congruent with their gender expression (Gerhardstein and Anderson 2010). This research suggests that, at least regarding appearance, people rate transgender individuals who conform to their gender identity, rather than their assigned birth sex, more positively. However, it is unknown whether this holds true for conformity to gender normative behavior and interests. Transgender individuals who conform to their gender identity may also be perceived as “passing” as cisgender more easily and so as more threatening. Gender identity “passing” may be viewed as more threatening for those higher in gender binarism beliefs for whom there exist only two genders (that conform to biological sex). For those higher in gender binarism beliefs, a “violator” of those gender boundaries would be threatening enough, but a “violator” who could not be detected easily may be perceived as even more of a threat.

The Current Studies

Past studies have investigated the role of gender conformity in relation to sexual orientation and perceptions of gay and lesbian targets (Blashill and Powlishta 2009, 2012). However, little research has experimentally investigated the role of gender conformity in relation to evaluating targets based on gender identity. Based on social identity theory and research on distinctiveness threat, it may be that anti-transgender prejudice stems from perceptions that transgender people transgress or blur boundaries between the binary genders. If so, transgender individuals who are highly gender role conforming to their gender expression may be more threatening because they demonstrate that gender is non-essential and that boundaries separating men and women are malleable or even non-existent. This effect may be particularly robust among those higher in gender binarism and gender essentialism because they believe that there are only two genders, which are biologically determined (Ching and Xu 2018; Tebbe and Moradi 2012). Additionally, people who hold traditional (versus egalitarian) gender role beliefs may be more threatened by gender conforming transgender people because traditional gender roles are based on an underlying ideology that gender differences are biologically based (versus socially constructed; Larsen and Long 1988). Anti-transgender attitudes may also influence how people view gender conformity among transgender people. For those higher in anti-transgender prejudice, a gender conforming transgender person may be more threatening because they are not detectable as targets of prejudice.

The current studies sought to disambiguate how gender conformity affects attitudes toward a transgender male or female target. Specifically, we assessed how attitudes toward a gender conforming versus gender nonconforming transgender person (Study 1: woman, Study 2: man, Study 3: man or woman) differ from attitudes toward a gender conforming versus nonconforming cisgender person (i.e., someone whose gender-identity conforms with their assigned birth sex; not transgender). In all three studies, we looked at conformity to the gender identity of the target (i.e., a feminine transgender woman and a feminine cisgender woman are both conforming to female gender roles).

Study 1 focused on attitudes toward transgender women; Study 2 focused on transgender men; and Study 3 directly compared attitudes toward transgender women and men. In all three studies, participants read a vignette depicting either a transgender or a cisgender target who presents as highly gender conforming or not gender conforming. Participants rated their liking of the target in the vignette, their acceptance of the target’s gender expression, and their level of distinctiveness threat. They also completed individual difference measures (i.e., anti-transgender prejudice in Studies 1 and 2, traditional gender roles in Studies 1 and 2, and gender essentialism in Study 3) that we predicted would be related to the dependent variables (i.e., liking, distinctiveness threat), and that those relationships would be moderated by the target’s gender identity and gender conformity. For all three studies, all materials and procedures were approved by an IRB.

Study 1: Transgender Prejudice Targeting Women

In our first study, we investigated attitudes toward women (cisgender and transgender). The design of Study 1 was a 2 (target gender identity: transgender woman or cisgender woman) × 2 (target gender-role conformity: feminine gender-conforming or feminine gender-nonconforming). Because transgender people are often the targets of prejudice and discrimination (Nagoshi et al. 2008; Tebbe and Moradi 2012), we hypothesized that participants would like transgender women less than cisgender women, regardless of target gender conformity (Hypothesis 1). Prior research has also demonstrated a consistent trend in cisgender men having more negative attitudes toward transgender people than cisgender women have (Antoszewski et al. 2007; Norton and Herek 2013; Tebbe and Moradi 2012). We hypothesized that male participants would have greater dislike of transgender targets than would female participants (Hypothesis 2).

Gender-based distinctiveness threat should be elicited by targets who transgress against the traditional gender roles associated with biologically-essential binary sexes. Thus, distinctiveness threat should be greater for the transgender female targets than for cisgender female targets (Hypothesis 3), as well as greatest for the gender-conforming transgender female target (Hypothesis 4). Pre-existing negative attitudes toward transgender people and rigid traditional gender role beliefs may activate negative responses to the targets based on gender identity and gender conformity.

We also predicted that the effect of anti-transgender prejudice (Hypothesis 5) and traditional gender role beliefs (Hypothesis 6) on ratings of the targets would be moderated by gender conformity and gender identity. Specifically we hypothesized that the relationships between target ratings on liking and distinctiveness threat and anti-transgender prejudice would be moderated by target gender identity (Hypothesis 5a) and by target gender conformity (Hypothesis 5b) such that the relationships between anti-transgender prejudice with both liking and distinctiveness threat would be more pronounced for transgender and nonconforming targets. We predicted the same pattern for the relationships between both liking and distinctiveness threat with traditional gender role beliefs (Hypothesis 6). Specifically, we hypothesized that the relationships between both liking and distinctiveness threat with traditional gender role beliefs would be more pronounced for transgender targets (Hypothesis 6a) and gender nonconforming targets (Hypothesis 6b).

Our predictions for Hypotheses 5 and 6 treat target gender identity and target gender conformity separately and independently by testing for simple moderation. Although we offer no formal hypotheses, we further wanted to explore the interaction of target gender identity and gender conformity, particularly focusing on the gender-conforming cisgender targets who are, by definition, the most normative group. To do so, we conducted moderated moderation analyses in which gender conformity served as the superordinate contextual moderator within which targets held either transgender or cisgender identities. Through four moderated moderation analyses, we were able to explore the relationships between both liking and distinctiveness threat with both anti-transgender prejudice and traditional gender role beliefs across the four target groups resulting from the crossing of gender identity (transgender and cisgender) with gender conformity (conforming and nonconforming). For all studies, we report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study.

Method

Participants

Participants were cisgender (as determined by responses to a single-item measure of gender identity where participants chose one of the following categories: male, female, non-cisgender [transgender, intersexed, or other – open text]), undergraduate psychology students at a U.S. Midwestern Jesuit University who received course credit for participation. Although 323 students participated (a priori power analysis indicated 231 were needed to obtain power = .90), 91 (28%) failed the reading or attention checks and were excluded from analyses. (Participants needed to correctly answer all three reading check questions and follow all of the Instructional Manipulation Check attention check instructions, described in the following section, to be included.) The final sample then included 232 students, with a mean age of 18.69 years-old (SD = 1.01, range = 18–23) and the majority of whom were female (152, 65.5%) and White (177, 76.3%; 32, 13.8%, Asian; 6, 2.6%, Black/Hispanic/Biracial). Most students were moderately to very religious (134, 57.7%) and Catholic (117, 50.4%). Many identified as moderate in their political orientation (94, 40.9%; 76, 33.1% liberal; 60, 26.1% conservative). A majority of students (145, 62.5%) reported having no contact with a transgender person (friend, family member, acquaintance) and 37.5% (n = 87) reported having some kind of contact with a transgender person. Male and female participants were comparable according to their race, χ²(5) = 8.53, p = .13, Cramer’s V = .19; religious affiliation, χ² (8) = 8.81, p = .36, Cramer’s V = .20; and religiosity, χ² (3) = 1.73, p = .63, Cramer’s V = .09. The one exception was political orientation, χ² (4) = 12.42, p = .01, Cramer’s V = .23, such that more women were liberal and more men were conservative.

Vignettes

Four vignettes (modified from Horn 2007) described “Jamie” as a female college student who displays traditional female gender-role conforming behavior or nonconforming behavior and was identified (with definitions) as either transgender or cisgender (bracketed phrases indicate the gender conformity manipulation and italicized bracketed phrases indicate the gender identity manipulation):

Jamie is a college student. Jamie plays on the school volleyball team. Jamie is a “B” student. Jamie dresses and acts like [differently from] most of the other girls at school. Jamie has long hair, and enjoys wearing make-up and dresses [a crew cut, and never wears make-up or dresses]. Jamie is biologically male (i.e., was born with a penis) and identifies as a transgender woman. [Jamie is biologically female (i.e., was born with a vagina) and identifies as a cisgender woman.]

There is a precedent for using the specific indicators presented in these vignettes to operationalize gender conformity and nonconformity (e.g., Horn 2007). The gender nonconforming vignette was taken directly from Horn (2007) and only modified by changing the target’s name and making the target a college student rather than a high school student. However, Horn’s gender conforming vignettes did not have as much detail about the target’s behavior and appearance (i.e., “dresses and acts like most of the other girls at school”). In order to make the vignettes more balanced, we added a description of behaviors and appearance to the gender conforming vignettes that were equivalent to those in Horn’s gender nonconforming vignettes.

Reading and Attention Checks

Following the vignette, participants answered three multiple-choice reading check questions: “What was the name of the person you read about?” “What was one of the hobbies of the person you read about?,” and “What gender identity was the person you read about?” The items ensured that participants read and understood the pertinent information in the vignettes. We also included Instructional Manipulation Check (IMC; Oppenheimer et al. 2009) questions within the questionnaires in the study. The IMC questions appear in identical format to the other items in the questionnaire but instruct participants to leave the item blank (i.e., not select any scale options). These items act as attention checks; participants failing the attention checks were excluded from analyses.

Dependent Variables: Liking and Distinctiveness Threat

We created questions for our dependent variables focused on (a) liking of the target and (b) distinctiveness threat elicited by the target. Three questions for liking (“I like Jamie,” “I would want to be friends with Jamie,” and “I would like to meet Jamie”) used a 7-point Likert-type scale ranging from 1 (not at all) to 7 (very much). We created a composite liking score across all three items (α = .92) by averaging across items such that higher scores indicate stronger liking. Two items (modified from Warner et al. 2007) rated on the same 7-point scale were used for distinctiveness threat: “Jamie blurs the boundary between female and male genders” and “Jamie threatens the differences between people who are born female and people who are born male.” A composite score was created for the distinctiveness threat items (corrected split-half r = .72) wherein higher averaged scores indicated stronger distinctiveness threat.

We also included a single multiple-choice item to examine the acceptance of gender expression: “What pronoun would you use to describe this person?” he/his; she/hers (coded as “correct” when the pronouns matched the gender expression of the target). However, in all three studies, the vast majority (91% in Study 1; 85.4% in Study 2; 74.4% in Study 3) reported that they would use the correct pronoun, creating a ceiling effect. As such, we excluded this variable from any further analyses.

Individual Difference Measures: Prejudice and Gender Role Attitudes

We included individual difference measures to be used in the moderated moderation analyses. The general attitudes subscale of the Prejudice toward Trans Women Scale (Winter et al. 2009) is a 17-item scale measuring attitudes toward transgender women (e.g., “Transwomen… Are men with something wrong in their mind”; “…Are in some way unnatural”), using a Likert-type scale from 1 (strongly disagree) to 5 (strongly agree). Higher averaged scores indicate greater anti-transgender prejudice. This scale showed good reliability in our sample (α = .91).

The Traditional-Egalitarian Sex Roles Scale (TESR; Larsen and Long 1988) is a 20-item scale assessing support for traditional versus egalitarian gender roles for both men (e.g., “Men who cry have weak character”) and women (e.g., “A woman’s place is in the home”). Responses are made to statements using a 5-point Likert-type scale from 1 (strongly disagree) to 5 (strongly agree). Higher averaged scores indicate greater adherence to traditional gender roles (and less acceptance of egalitarian gender roles). The scale showed good reliability in our sample (α = .85).

Procedure

Participants completed the study online. Participants were randomly assigned to read one of the four vignettes describing Jamie: feminine-conforming transgender woman, nonconforming transgender woman, feminine-conforming cisgender woman, or nonconforming cisgender woman. Following the vignette, participants were provided with definitions of key terms (i.e., “A transgender woman is someone who is assigned male at birth but identifies and lives as a woman;” “A cisgender woman is someone whose gender-identity conforms with their biological sex; not transgender”) and instructed to keep these definitions in mind when completing the rest of the questionnaires. Participants rated their liking of Jamie, how accepting they were of Jamie’s gender expression, and their level of distinctiveness threat.

Participants also completed questionnaires of prejudice toward transgender women, traditional-egalitarian gender role preference, and demographic information (i.e., age; gender identity; ethnicity; if they are United States citizens; if English is their first language; approximate population of their hometown; zip code of their hometown; political orientation; political party affiliation; political leanings on social, economic, and foreign policy issues; religion; religiosity; contact with transgender people [family member, friend, acquaintance]). Participants also completed a measure of social distance from transgender women (i.e., modified from Gentry 1987); however, no significant results were found for social distance and the results pertaining to this measure are not included.

Results

ANOVAs

To assess liking and distinctiveness threat elicited by the targets, we conducted a 2 (target gender identity: transgender woman or cisgender woman) × 2 (target gender-role conformity: feminine gender-conforming or feminine gender-nonconforming) × 2 (participants’ binary gender: male or female) ANOVA for each dependent variable: liking and distinctiveness threat.

Liking

We found a main effect of participant gender on liking of the targets, F(1,222) = 21.82, p < .001, η _p² = .09; men (M = 4.21, SD = 1.28) liked the targets less than did women (M = 5.11, SD = 1.22). There was also a main effect of target gender identity, F(1,222) = 6.94, p = .009, η _p² = .03; as predicted, cisgender targets (M = 5.01, SD = 1.20) were liked more than transgender targets (M = 4.63, SD = 1.38) regardless of gender conformity, supporting Hypothesis 1 which predicted that transgender targets would be liked less than cisgender targets.

There was also a two-way interaction between participants’ gender and target gender identity, F(1,222) = 8.90, p = .003, η _p² = .04. Simple effects tests indicated that the effect was significant only for transgender targets, F(1,119) = 32.70, p = .001, η _p² = .22, not for cisgender targets, F(1,107) = .1.28, p = .26. Male participants’ liking of the transgender targets (M = 3.84, SD = 1.18) was significantly lower than was female participants’ liking of the transgender targets (M = 5.14, SD = 1.26), supporting Hypothesis 2, which predicted male participants would like transgender targets less than female participants. None of the remaining main effects, two-way, nor three-way interactions was significant.

Because political orientation differed between male and female participants, we also conducted an ANCOVA with political orientation as the covariate to test that the effects of participants’ gender were not actually artifacts of the significantly different political orientations among the binary participant gender groups. We found that the effects of participants’ gender remained significant even when controlling for political orientation.

Distinctiveness Threat

We found a main effect of target gender conformity, F(1,222) = 7.45, p = .007, η _p² = .03; feminine-conforming targets (M = 2.20, SD = 1.52) elicited less distinctiveness threat than did feminine nonconforming targets (M = 2.86, SD = 1.59). There was also a main effect of target gender identity, F(1,222) = 40.84, p < .001, η _p² = .16; cisgender women (M = 1.86, SD = 1.28) elicited less distinctiveness threat than did transgender women (M = 3.17, SD = 1.58), supporting Hypothesis 3 which predicted that distinctiveness threat would be higher for transgender than for cisgender targets. There were no significant interactions, thus failing to support Hypothesis 4, which predicted that conforming transgender targets would generate the highest levels of distinctiveness threat. These effects also remained unchanged when we conducted an ANCOVA analysis that included political orientation as a covariate.

Moderated Moderation Analyses

To begin, bivariate correlations among the variables in the moderated moderation models suggest small-to-medium relationships among the variables, suggesting little to no multicollinearity (see Table 1). The initial patterns show a significant positive correlation between anti-transgender prejudice and traditional gender role beliefs. There were also significant positive correlations between distinctiveness threat with target conformity, target gender identity (such that distinctiveness threat is higher for nonconforming targets and transgender targets), anti-transgender prejudice, and traditional gender role beliefs. Liking was significantly negatively correlated with target gender identity (such that liking was lower for transgender targets), anti-transgender prejudice, traditional gender role beliefs, and distinctiveness threat.

Table 1 Correlations among variables in the regression models, Studies 1 and 2

We hypothesized that the effect of anti-transgender prejudice (Hypothesis 5) and traditional-egalitarian gender role beliefs (Hypothesis 6) on ratings of the target would be moderated by gender identity (Hypotheses 5a and 6a) and gender conformity (Hypotheses 5b and 6b). We predicted that the moderators (i.e., target gender identity, target gender conformity) each would have a strengthening effect on the dependent variables, such that the relationships between the individual difference variables (i.e., anti-transgender prejudice, traditional gender role beliefs) and both liking and distinctiveness threat would be more pronounced for transgender and nonconforming targets (i.e., simple moderation). We also explored a possible moderated moderation relationship between target gender identity and gender conformity such that the relationships between anti-transgender prejudice/traditional gender role beliefs with both liking and distinctiveness threat would be differentiated for transgender and cisgender targets based on their gender conformity. We conducted four independent moderated moderation analyses to assess the combined and separate influence of target gender identity and target gender conformity on the relationships between (a) liking with anti-transgender prejudice, (b) liking with traditional gender role beliefs, (c) distinctiveness threat with prejudice, and (d) distinctiveness threat with gender role beliefs.

All analyses were conducted using PROCESS Model 3 (Hayes 2013). Anti-transgender prejudice and traditional gender role beliefs were centered. Target conformity was coded as 0 = conforming and 1 = nonconforming. Gender identity was coded such that 0 = cisgender and 1 = transgender. We chose not to include gender of the participant in the regression models because it would have added an additional moderator, which would be difficult to interpret.

Liking

Our first moderated moderation analysis examined the effects of target conformity and gender identity on the relationship between anti-transgender prejudice and liking. The moderators were target gender conformity and target gender identity, the predictor was anti-transgender prejudice, and the criterion was liking. The overall model was significant (R² = .37, MSE = 1.11), F(7,223) = 18.46, p < .001. Both gender identity (β = −.54, SE = .21, t = −2.64, p = .009) and gender conformity (β = −.54, SE = .21, t = −2.59, p = .01) were significant predictors of liking; as shown by the ANOVAs, liking was lower for transgender and nonconforming targets. The relationship between liking and transgender prejudice was significantly moderated by gender identity (β = −1.23, SE = .30, t = −4.04, p < .001) and by gender conformity (β = −.73, SE = .31, t = −2.38, p = .02) (i.e., both simple moderation analyses were significant). As anti-transgender prejudice increased, liking of transgender targets and of nonconforming targets decreased, supporting Hypotheses 5a and 5b which predicted that transgender (Hypothesis 5a) and nonconforming (Hypothesis 5b) targets would be liked less by those scoring higher in anti-transgender prejudice.

The overall moderated moderation analysis predicting liking was also significant (β = 1.07, SE = .43, t = 2.48, p = .01) and increased the amount of explained variance (∆R² = .02), F(1, 223) = 6.13, p = .01. To probe this model, we looked at simple moderation of target gender identity within each of the conformity conditions. For conforming targets, the simple moderated model was significant (β = −1.23, SE = .30, t = −4.04, p < .001; see Fig. 1a), but it was not significant for nonconforming targets (β = −.16, SE = .30, t = −.54, p = .59; see Fig. 1b). Simple slopes analyses showed that higher prejudice was associated with lower liking for conforming (β = −1.60, SE = .23, t = −6.93, p < .001) and nonconforming transgender (β = −1.27, SE = .19, t = −6.60, p < .001) as well as nonconforming cisgender (β = −1.10, SE = .24, t = −4.67, p < .001) targets. However, prejudice and liking were unrelated for conforming, cisgender targets (β = −.37, SE = .20, t = −1.85, p = .07; see Fig. 1).

Fig. 1

Moderation effect of target gender identity and target gender conformity on the relationship between anti-transgender prejudice and ratings of liking, study 1. Black lines (both solid and dashed) indicate a significant negative relationship between prejudice and liking; the grey, dashed line indicates a nonsignificant effect for the conforming, cisgender target

Our second moderated moderation analysis tested the effects of target conformity and gender identity on the relationship between traditional gender role beliefs and liking. The model included target gender conformity and identity as moderators, gender role beliefs as the predictor, and liking as the criterion. The overall model was significant (R² = .24, MSE = 1.34), F(7,223) = 10.32, p < .001. Target gender identity (β = −.53, SE = .22, t = −2.40, p = .02) was a significant predictor of liking, showing, consistent with the ANOVA, that transgender targets were liked less. For the simple moderator analyses, the relationship between traditional gender role beliefs and liking was moderated by target gender identity (β = −1.60, SE = .46, t = −3.49, p < .001), such that as traditional gender role beliefs increased, liking of transgender targets decreased, supporting Hypothesis 6a which predicted that liking of transgender targets would be lower among those higher in traditional gender role beliefs. Hypothesis 6b (i.e., liking of nonconforming targets would be lower among those higher in traditional gender role beliefs) was not supported (β = −.65, SE = .46, t = −1.42, p = .16). Finally, the test of moderated moderation for the second model was not significant (β = .95, SE = .65, t = −1.46, p = .14).

Distinctiveness Threat

Our third moderated moderation analysis assessed the effect of target conformity and gender identity on the relationship between anti-transgender prejudice and distinctiveness threat. The moderators were target gender conformity and target gender identity, the predictor was anti-transgender prejudice, and the criterion was distinctiveness threat. The overall model for the effects of target conformity, target gender identity, and anti-transgender prejudice on distinctiveness threat was significant (β = .43, MSE = 1.48), F(7,223) = 23.89, p < .001. Both gender identity (β = 1.63, SE = .24, t = 6.89, p < .001) and conformity (β = .99, SE = .24, t = 4.14, p < .001) were significant predictors of distinctiveness threat, such that transgender and nonconforming targets elicited more distinctiveness threat, which was consistent with the ANOVAs. Both simple moderation analyses were not significant. The relationship between anti-transgender prejudice and distinctiveness threat was not moderated by target gender identity (β = .66, SE = .35, t = 1.89, p = .06) or by target gender conformity (β = .11, SE = .36, t = .32, p = .75). Thus, neither Hypothesis 5a (i.e., transgender would elicit greater distinctiveness threat for those scoring higher in anti-transgender prejudice) nor Hypothesis 5b (i.e., nonconforming targets would elicit greater distinctiveness threat for those scoring higher in anti-transgender prejudice) were supported. The test of moderated moderation for the third model was also not significant (β = .11, SE = .50, t = .22, p = .83).

The fourth moderated moderation analysis assessed the effect of target conformity and gender identity on the relationship between traditional gender role beliefs and distinctiveness threat. The moderators were target gender conformity and target gender identity, the predictor was traditional gender role beliefs, and the criterion was distinctiveness threat. This model was also significant (β = .29, MSE = 1.91), F(7,223) = 12.82, p < .001). Both gender identity (β = 1.64, SE = .27, t = 6.17, p < .001) and conformity (β = .80, SE = .27, t = 3.01, p = .003) significantly predicted distinctiveness threat, such that distinctiveness threat was greater for transgender and for gender nonconforming targets, which was consistent with the ANOVA. The simple moderation analyses were not significant. The relationship between traditional gender role beliefs and distinctiveness threat was not moderated either by target gender identity (β = .76, SE = .55, t = 1.38, p = .17) or by target gender conformity (β = −.20, SE = .58, t = −.34, p = .74). Thus, Hypothesis 6a (i.e., transgender targets would elicit greater distinctiveness threat for those scoring higher in traditional gender role beliefs) and Hypothesis 6b (i.e., gender nonconforming targets would elicit greater distinctiveness threat for those scoring higher in traditional gender role beliefs) were not supported. The test of moderated moderation for the fourth model was also not significant (β = −.45, SE = .78, t = −.58, p = .56).

Discussion

The present study investigated attitudes toward cisgender and transgender women who were gender conforming or nonconforming. We found that regardless of gender conformity, transgender targets were liked less than cisgender targets and that distinctiveness threat was greater for the transgender targets (compared to cisgender targets) and nonconforming targets (compared to conforming targets). Additionally, the relationship between liking and anti-transgender prejudice was moderated by both target gender identity and target gender conformity, such that liking decreased as scores in anti-transgender prejudice increased when reading about a conforming transgender target or a nonconforming target of either gender identity.

Consistent with prior research (e.g., Norton and Herek 2013), male participants expressed more negative attitudes toward transgender targets. Transgender targets were also liked less and elicited more distinctiveness threat than cisgender targets did, which is consistent with prior research on attitudes toward transgender individuals (Tebbe and Moradi 2012). Additionally, gender nonconforming targets elicited more distinctiveness threat. We found that transgender women and nonconforming women are viewed more negatively and are perceived as transgressing the distinct boundaries between binary genders. These findings suggest that people still tend to view assigned birth sex as primary and transgender women as transgressing their assigned birth sex (i.e., “men that dress as women;” Gazzola and Morrison 2014, p. 83). Additionally, gender nonconforming women (regardless of gender identity) also elicited distinctiveness threat, suggesting that any woman who appears or acts in a traditionally masculine manner is also perceived as threatening the distinction between binary genders. However, we did not find an interaction between gender identity and gender conformity, which may suggest that, at least for transgender women, conformity to traditional gender roles for their gender expression does not affect people’s perceptions of them.

Study 2: Transgender Prejudice Targeting Men

Based on the results of Study 1, we were interested in whether these results would be replicated with a transgender male target. Most research has not separately examined attitudes toward transgender women and transgender men (for some exceptions see Gerhardstein and Anderson 2010; Nagoshi et al. 2018), but it has instead focused on attitudes toward transgender people more generally. The design of Study 2 was the same as Study 1, but with cisgender and transgender men as the targets instead of women, that is a 2 (target gender identity: transgender man or cisgender man) × 2 (target gender-role conformity: masculine gender-conforming or masculine gender-nonconforming) design. We made the same hypotheses (but for male targets) as in Study 1: We hypothesized that participants would like male transgender targets less than male cisgender targets (Hypothesis 1) and that male participants would like male transgender targets less than female participants (Hypothesis 2). We also hypothesized that male transgender targets would elicit greater distinctiveness threat than male cisgender targets (Hypothesis 3) and that gender-conforming transgender targets would elicit the greatest amount of distinctiveness threat (Hypothesis 4). We again predicted that target gender identity and target gender conformity would moderate the relationships between the individual difference variables (i.e., anti-transgender prejudice, traditional gender role beliefs) and the dependent variables (i.e., liking, distinctiveness threat). Specifically, we again predicted that that liking would be less and distinctiveness threat would be greater for those scoring higher in anti-transgender prejudice when rating transgender targets (Hypothesis 5a) and nonconforming targets (Hypothesis 5b). Similarly, we predicted that liking would be less and distinctiveness threat would be greater for those scoring higher in traditional gender role beliefs when rating transgender targets (Hypothesis 6a) and nonconforming targets (Hypothesis 6b).

Method

Participants

Participants were undergraduate introductory psychology students at a U.S. Midwestern Jesuit University who completed the study online and received course credit for their participation. Fully 282 students participated (a power analysis indicated 231 were needed to obtain power = .90). However, 61 (21.6%) participants were excluded for failing reading and attention checks (using the same exclusion criteria as in Study 1) and four (1.4%) were excluded for self-identifying as non-cisgender (on a single-item gender identity item), because four is not enough to analyze separately, leaving 217 (observed power = .92) viable respondents. The demographics were very similar to those of the Study 1 sample: young (M_age = 19.06, SD = 1.12, range = 18–24), mostly female (n = 144, 66.4%), White (162, 74.7%; 30, 13.8% Asian; 9, 4.1% Hispanic/Latino; 6, 2.8% Black; 10, 4.6% Biracial/other), and moderately to very religious (125, 57.6%) and Catholic (115, 53%). The majority identified as moderate in their political orientation (105, 48.4; 63, 29.1% liberal; 50, 22.3% conservative). Only slightly more students did not have contact with a transgender person (119, 54.8%) than did have contact with a transgender person (98, 45.2%). Male and female participants were comparable demographically: race, χ²(5) = 6.31, p = .28, Cramer’s V = .17; political orientation: χ²(4) = 3.14, p = .53, Cramer’s V = .12, with the exception of religious affiliation,χ² (10) = 20.48, p = .03, Cramer’s V = .31, and religiosity, χ² (3) = 11.18, p = .01, Cramer’s V = .23. More women identified as Catholic and as moderately religious than did men.

Materials

We replicated the materials used in Study 1, with the exception that we used the male gender nonconforming vignettes from Horn (2007). The male vignettes were modified in the same way as the female vignettes in Study 1 (i.e., target name and school level altered; equivalent gender-conforming behaviors and appearance descriptions added). The male vignettes depict Jamie as a transgender or cisgender man who dresses and acts in a traditionally masculine manner (i.e., gender-conforming) or in a more traditionally feminine manner (i.e., gender-nonconforming). Bracketed phrases indicate the gender conformity manipulation and italicized bracketed phrases indicate the gender identity manipulation:

Jamie is a college student. Jamie plays on the school baseball team. Jamie is a “B” student. Jamie dresses and acts like [differently from] most of the other guys at school. Jamie acts masculine, has short hair, and mostly wears jeans and t-shirts [feminine, and sometimes wears fingernail polish and eyeliner]. Jamie is biologically male (i.e., was born with a penis) and identifies as a cisgender man. [Jamie is biologically female (i.e., was born with a vagina) and identifies as a transgender man.]

Throughout the survey, terms and pronouns were altered to reflect transgender men (versus transgender women, as in Study 1). As is Study 1, we had acceptable reliabilities for all of the measures: liking (α = .94), distinctiveness threat (corrected split-half r = .71), anti-transgender prejudice (α = .87), and traditional gender roles (α = .83).

Procedure

The procedure was identical to that of Study 1. The definitions provided were for cisgender and transgender men: “A transgender man is someone who is assigned female at birth but identifies and lives as a man”; “A cisgender man is someone whose gender-identity conforms with their biological sex; not transgender.”

Results

ANOVAs

We conducted 2 (target gender identity: transgender man or cisgender man) × 2 (target gender-role conformity: masculine gender-conforming or masculine gender-nonconforming) × 2 (participants’ binary gender: male or female) ANOVAs to assess liking and distinctiveness threat for the targets.

Liking

As in Study 1, there was a main effect of participants’ gender on liking, F(1,209) = 8.37, p = .004, η _p² = .04, where men (M = 4.21, SD = 1.39) liked all of the targets less than women (M = 4.80, SD = 1.32) did, partially supporting Hypothesis 2, which predicted that male participants would like transgender targets less than would female participants. There were no main effects for target gender identity, which failed to support Hypothesis 1, which predicted transgender targets would be liked less than cisgender targets. There were no other significant main effects or interactions for liking.

Distinctiveness Threat

There were main effects of target gender conformity, F(1,209) = 25.89, p < .001, η _p² = .11, and of target gender identity, F(1,209) = 31.38, p < .001, η _p² = .13. Gender nonconforming targets (M = 3.05, SD = 1.35) elicited greater distinctiveness threat than did conforming targets (M = 2.04, SD = 1.22), and transgender targets (M = 2.96, SD = 1.41) elicited more distinctiveness threat than did cisgender targets (M = 2.10, SD = 1.20), supporting Hypothesis 3, predicting transgender targets would elicit greater distinctiveness threat than would cisgender targets. There was also an interaction of target gender identity and target gender conformity, F(1,209) = 10.00, p = .002, η _p² = .05. Simple effects tests indicated that gender conforming transgender targets (M = 2.69, SD = 1.25) elicited greater distinctiveness threat than did gender conforming cisgender targets (M = 1.36, SD = .71), F(1,103) = 58.03, p < .001, η _p² = .36, whereas there was not a significant difference in distinctiveness threat posed by nonconforming targets based on gender identity, F(1,106) = 2.19, p = .14. These results support Hypothesis 4, which predicted conforming transgender targets would elicit the greatest distinctiveness threat. There were no other significant main effects or interactions.

Because religiosity differed between male and female participants, we also conducted ANCOVAs with religiosity as the covariate for both the analyses of liking and distinctiveness threat. We found that the effects of participants’ gender remained significant even when controlling for religiosity.

Moderated Moderation Analyses

As in Study 1, we hypothesized that target gender identity and target gender conformity would moderate the relationship between anti-transgender prejudice and traditional gender role beliefs and both liking and distinctiveness threat, such that transgender and nonconforming targets would be liked less and elicit greater distinctiveness threat among those scoring higher in anti-transgender prejudice or in traditional gender role beliefs. We conducted the same set of four moderated moderation analyses as in Study 1. Anti-transgender prejudice and traditional gender role beliefs were centered. Target conformity was coded as 0 = conforming and 1 = nonconforming. Gender identity was coded such that 0 = cisgender and 1 = transgender. All analyses were conducted using PROCESS Model 3 (Hayes 2013).

We conducted correlations among the variables included in the models and found small-to-moderate relationships among the variables (see Table 1). As in Study 1, anti-transgender prejudice and traditional gender role beliefs were positively correlated. Distinctiveness threat was again significantly positively correlated with target conformity and target gender identity (i.e., higher for nonconforming targets and transgender targets) and anti-transgender prejudice. Liking was significantly negatively correlated with anti-transgender prejudice, traditional gender role beliefs, and distinctiveness threat, but not with target characteristics. As in Study 1, we did not include gender of the participant in the regression models for parsimony and because no effects of participants’ gender were found in the ANOVAs.

Liking

The first moderated moderation analysis, testing the moderation of anti-transgender prejudice on liking by target gender conformity and target gender identity, was significant (R² = .36, MSE = 1.24), F(7,209) = 17.08, p < .001. Target gender identity and target gender conformity were not significant predictors of liking, consistent with the ANOVA findings. The relationship between liking and anti-transgender prejudice was significantly moderated by target gender conformity (β = −1.39, SE = .38, t = −3.65, p < .001) as well as by target gender identity (β = −1.91, SE = .38, t = −5.07, p < .001), such that as anti-transgender prejudice increased, liking decreased for gender nonconforming targets and for transgender targets. These patterns support Hypothesis 5a (i.e., transgender targets would be liked less by those scoring higher in anti-transgender prejudice) and Hypothesis 5b (i.e., nonconforming targets would be liked less by those scoring higher in anti-transgender prejudice).

The overall moderated moderation analysis predicting liking was also significant (β = 1.55, SE = .52, t = 2.99, p < .001; see Fig. 2) and increased the amount of variance explained (∆R² = .03), F(1, 209) = 8.94, p < .001. To probe this model, we looked at the simple moderation of target gender identity within each of the conformity conditions. For conforming targets, the simple moderated model was significant (β = −1.91, SE = .38, t = −5.07, p < .001; see Fig. 2a), but it was not significant for nonconforming targets (β = −.36, SE = .36, t = −1.01, p = .31; see Fig. 2b). Simple slopes analyses showed that higher prejudice was associated with lower liking for conforming (β = −1.75, SE = .29, t = −6.11, p < .001) and nonconforming (β = −1.59, SE = .21, t = −7.69, p < .001) transgender as well as nonconforming cisgender (β = −1.23, SE = .29, t = −4.21, p < .001) targets. However, prejudice and liking were unrelated for conforming, cisgender targets (β = .16, SE = .25, t = .67, p = .51; see Fig. 2).

Fig. 2

Moderation effect of target gender identity and target gender conformity on the relationship between anti-transgender prejudice and ratings of liking, Study 2. Black lines (both solid and dashed) indicate a significant negative relationship between prejudice and liking; the grey, dashed line indicates a nonsignificant effect for the conforming, cisgender target

The second moderated moderation analysis tested the moderation of traditional gender role beliefs on liking by target gender conformity and target gender identity. The overall model was significant (R² = .20, MSE = 1.55), F(7,209) = 7.52, p < .001. Target gender identity and target gender conformity were not significant predictors of liking, as in the ANOVA. The relationship between gender role beliefs and liking was significantly moderated by both target gender conformity (β = −1.34, SE = .64, t = −2.10, p = .04) and target gender identity (β = −1.35, SE = .57, t = −2.38, p = .02). As traditional gender role beliefs increased, liking decreased for nonconforming targets and for transgender targets, supporting Hypothesis 6a (i.e., transgender targets would be liked less by those scoring higher in traditional gender role beliefs) and Hypothesis 6b (i.e., nonconforming targets would be liked less by those scoring higher in traditional gender role beliefs). The test of moderated moderation was not significant (β = 1.35, SE = .84, t = 1.60, p = .11).

Distinctiveness Threat

The third moderated moderation analysis, testing the moderation of anti-transgender prejudice on distinctiveness threat by target gender conformity and target gender identity, was significant (R² = .35, MSE = 1.27), F(7,209) = 16.28, p < .001. Both target gender identity (β = 1.40, SE = .22, t = 6.42, p < .001) and target gender conformity (β = 1.55, SE = .23, t = 6.87, p < .001) were significant predictors of distinctivness threat, such that distinctiveness threat was greater for transgender and nonconforming targets. The relationship between anti-transgender prejudice and distinctiveness threat was significantly moderated by target gender identity (β = .82, SE = .38, t = 2.15, p = .03), such that as anti-transgender prejudice increased, distinctiveness threat increased for transgender targets. These results support Hypothesis 5a, which predicted that transgender targets would elicit greater distinctiveness threat for those scoring higher in anti-transgender prejudice. Target gender conformity did not moderate the relationship between anti-transgender prejudice and distinctiveness threat (β = .02, SE = .39, t = .04, p = .97), which did not support Hypothesis 5b (i.e., nonconforming targets would elicit greater distinctiveness threat for those scoring higher in anti-transgender prejudice). The test of moderated moderation was not significant (β = −.28, SE = .53, t = −.52, p = .60).

The final moderated moderation analysis, testing the moderation of gender role beliefs on distinctiveness threat by target gender conformity and target gender identity, was significant (R² = .28, MSE = 1.41), F(7,209) = 11.59, p < .001. Both target gender identity (β = 1.36, SE = .23, t = 5.90, p < .001) and target gender conformity (β = 1.59, SE = .24, t = 6.58, p < .001) were significant predictors of distinctiveness threat, such that distinctiveness threat was greater for transgender and nonconforming targets. The relationship between traditional gender role beliefs and distinctiveness threat was not moderated by either target gender identity (β = .42, SE = .54, t = .78, p = .44) or target gender conformity (β = .40, SE = .61, t = .66, p = .51); thus, Hypotheses 6a and 6b were not supported. The test of moderated moderation was not significant (β = −.61, SE = .80, t = −.76, p = .45).

Discussion

Our results for the male targets largely replicated those found in Study 1 with female targets. Transgender and gender nonconforming targets elicited greater distinctiveness threat than did cisgender or gender conforming targets. Additionally, for the male targets, we found an interaction between gender identity and gender conformity: Distinctiveness threat was higher for gender conforming transgender targets than for gender conforming cisgender targets. We again found that target gender identity and target conformity moderated the relationships between anti-transgender prejudice and traditional gender role beliefs on both liking and distinctiveness threat. As in Study 1, as anti-transgender prejudice increased, liking decreased for both conforming transgender targets and nonconforming targets (of both gender identities). Our second study demonstrates that overall, transgender men are liked less than cisgender men and that feminine men are liked less than masculine men.

Additionally, the relationship between liking and anti-transgender prejudice was moderated by target gender identity and target gender conformity, such that liking for transgender and nonconforming targets decreased as anti-transgender prejudice increased. In contrast to Study 1, we did find an interaction between gender identity and gender conformity for the male targets on distinctiveness threat. This supports our hypothesis that transgender individuals who conform to the gender roles associated with their gender expression may be perceived as a greater threat to binary gender distinctiveness, potentially because they would be harder to detect as “transgressors” (i.e., are passing). It is also possible that gender conforming transgender individuals threaten beliefs about the immutability of binary gender because they are “passing,” consistent with their gender expression but not with their birth sex. Because we did not find this predicted effect in Study 1 with women targets, we conducted a third study to examine both male and female targets together.

Study 3: Transgender Prejudice Targeting Women and Men

Although the results of Studies 1 and 2 were very similar, a key difference is that in Study 2, an interaction was found between target men’s gender identity and gender conformity (i.e., that distinctiveness threat was significantly higher for conforming transgender men than conforming cisgender men), but the same interaction was not found in Study 1 for women targets. Based on these differential findings, we were interested whether there is a difference in attitudes toward transgender men and transgender women, and if, specifically, gender conformity differentially impacts attitudes toward transgender men and transgender women. National statistics indicate that transgender women are more frequently the victims of anti-transgender discrimination and violence than transgender men are (James et al. 2016), which might suggest that attitudes are more negative toward transgender women than toward transgender men. However, previous research also suggests that attitudes toward gender nonconforming men are more negative than attitudes toward gender nonconforming women (McCreary 1994; Sirin et al. 2004), which might explain the differential effects of gender conformity for the women targets (Study 1) and men targets (Study 2). Conversely, among respondents to the 2008 National Transgender Discrimination survey, it was found that transgender women were more likely to be perceived as gender nonconforming than transgender men were (Miller and Grollman 2015), suggesting that there may be an interaction between gender identity and gender conformity for women targets that was not detected in Study 1. For these reasons, we sought to directly compare attitudes toward transgender and cisgender men and women depending on their gender conformity. Additionally, Study 3 sought to replicate the findings of Studies 1 and 2 with a larger, U.S. community sample (versus the U.S. university student samples utilized in the previous studies).

The design of Study 3 was a 2 (target gender identity: transgender or cisgender) × 2 (target gender-role conformity: gender-conforming or gender-nonconforming) × 2 (target gender expression: woman or man). We hypothesized that our results from Studies 1 and 2 would be replicated such that participants would like transgender targets less than cisgender targets and that transgender targets would elicit greater distinctiveness threat than cisgender targets, regardless of target gender conformity (Hypothesis 1). We also hypothesized that nonconforming male targets (both cisgender and transgender) would be liked less and elicit greater distinctiveness threat than nonconforming female targets (Hypothesis 2). We also predicted that (as shown in Study 2) distinctiveness threat would be greater for conforming transgender targets than for nonconforming transgender targets (Hypothesis 3). For parsimony, we chose to measure only a single potential individual difference variable that we had not previously assessed: gender essentialist beliefs. One potential cause for greater distinctiveness threat being elicited for conforming transgender targets in Study 2 is that “passing” transgender individuals challenge the idea that gender roles are directly linked to biological sex differences. Thus, we hypothesized that target gender identity (Hypothesis 4a) and target conformity (Hypothesis 4b) would moderate the relationship between gender essentialism and both liking and distinctiveness threat, such that liking would decrease and distinctiveness threat would increase as gender essentialist beliefs increased for both transgender and nonconforming targets.

Method

Participants

U.S. participants were recruited from Amazon’s Mechanical Turk (mTurk), a micro-worker site where adult workers complete online tasks posted by researchers and businesses for small sums of money. Fully 622 workers participated, although a large number (n = 155, 24.9%) failed to complete the survey or were excluded for failing reading and attention checks (using the same exclusion criteria as in Studies 1 and 2), and five participants (.8%) who identified as non-cisgender (based on self-identification on a single-item gender identity question) were excluded (because five is not enough to analyze separately), leaving 462 viable participants (a power analysis indicated that 357 participants were needed to obtain power = .95).

The mean age of participants was 38.57 (SD = 12.84, range = 18–75) and the majority were female (n = 301, 66.5%). A majority were White (366, 79.6%; 39, 8.5% Black; 26, 5.7% Asian; 16, 3.5% Hispanic/Latino; 2, .4% Native American; 11, 2.4% Biracial/other), liberal (188, 40.9%; 154, 33.6% moderate; 117, 25.5% conservative), and not at all to only a little religious (300, 64.9%). About an equal number did not have contact with a transgender person (222, 48.1%) as did have contact with a transgender person (240, 51.9%). Male and female participants were nearly comparable demographically: race, χ²(6) = 3.87, p = .70, Cramer’s V = .09, and political orientation, χ²(4) = 5.17, p = .22, Cramer’s V = .11, with the exceptions of religious affiliation, χ² (10) = 24.40, p = .007, Cramer’s V = .23, and religiosity, χ² (3) = 17.67, p = .001, Cramer’s V = .20. More women identified as Protestant than men, and more men identified as agnostic or atheist than did women. More women identified as “very religious,” and more men identified as “not at all religious.”

Dependent Variables: Liking and Distinctiveness Threat

We used the same questions about liking and distinctiveness threat. We had adequate-to-good reliability for the liking (α = .92) and distinctiveness threat (corrected split-half r = .69) measures in the present sample.

Biological Gender Essentialism

For the sake of parsimony, we replaced the two individual difference measures used previously with one new measure: Biological Gender Essentialism (Brescoll et al. 2013). This 7-item scale measures beliefs about the biological basis for gender differences (i.e., gender essentialism), using a Likert-type scale from 1 (completely disagree) to 11 (completely agree). Higher averaged scores indicate greater belief in biological gender essentialism. This scale showed good reliability (α = .88). Sample item: “I think that differences between men and women in behavior are largely determined by the biological differences between the genders.”

Procedure

We used the same vignettes as in Studies 1 and 2, such that participants were randomly assigned to read one of the eight vignettes—factorially crossing target’s gender expression (male, female), gender conformity (conforming. nonconforming), and gender identity (transgender, cisgender). Participants were provided with the same definitions for cisgender man, transgender man, cisgender woman, transgender woman as used in Studies 1 and 2. Participants then completed the dependent and individual difference measures and the same demographic questions as in Study 1.

Results

ANOVAs

To test our hypotheses that participants would report less liking and greater distinctiveness threat for transgender targets than cisgender targets (Hypothesis 1), that nonconforming male targets (regardless of gender identity) would be liked less and elicit greater distinctiveness threat than women targets (Hypothesis 2), and that conforming transgender targets would elicit more distinctiveness threat than nonconforming transgender targets (Hypothesis 3), we conducted an ANOVA for each dependent variable. Unlike Studies 1 and 2, the design of Study 3 was a 2 (target gender identity: transgender or cisgender) × 2 (target gender-role conformity: gender-conforming or gender-nonconforming) × 2 (target gender expression: female or male); for parsimony, we did not include participants’ gender in the ANOVA (i.e., keeping it a 2 × 2 × 2).

Liking

We found a main effect for target gender expression, F(1,454) = 4.96, p = .03, η _p² = .01, where male targets (M = 4.28, SD = 1.51) were liked less than female targets (M = 4.58, SD = 1.37). There was also an interaction of target gender identity and target gender expression, F(1,454) = 5.54, p = .02, η _p² = .02. Simple effects indicated that there was a significant difference in liking of female targets only, F(1,231) = 5.57, p = .02, η _p² = .02, such that cisgender female targets (M = 4.79, SD = 1.29) were liked more than transgender female targets (M = 4.37, SD = 1.41), partially supporting Hypothesis 1 that transgender targets would be liked less than cisgender targets. There was not a significant difference in liking among male targets based on gender identity, F(1,227) = 1.12, p = .29. No other main effects or interactions were significant, failing to fully support Hypothesis 1 (i.e., transgender targets, regardless of gender expression, would be liked less than cisgender targets). Hypothesis 2 was also unsupported: We did not find evidence that nonconforming male targets (M = 4.31, SD = 1.57) were liked less than nonconforming female targets (M = 4.62, SD = 1.32).

Distinctiveness Threat

There was a main effect of target conformity, F(1,454) = 16.61, p < .001, η _p² = .04, and a main effect of target gender identity, F(1,454) = 59.17, p < .001, η _p² = .12, on distinctiveness threat. Nonconforming targets elicited greater distinctiveness threat (M = 2.93, SD = 1.48) than did conforming targets (M = 2.36, SD = 1.73), and transgender targets elicited more distinctiveness threat (M = 3.18, SD = 1.61) than did cisgender targets (M = 2.11, SD = 1.47), supporting Hypothesis 1 which predicted transgender targets would elicit greater distinctiveness threat than cisgender targets.

There was also an interaction between target conformity and target gender identity, F(1,454) = 33.79, p < .001, η _p² = .07. Simple effects tests indicated there was not a significant difference in distinctiveness threat between conforming transgender targets and nonconforming transgender targets, F(1,228) = 1.35, p = .25, which fails to support Hypothesis 3 predicting that conforming transgender targets would elicit greater distinctiveness threat than would nonconforming transgender targets. However, looking at the simple effects for conforming targets and nonconforming targets, the effect was significant among conforming targets only, F(1,233) = 93.14, p < .001, η _p² = .29; for nonconforming targets, the effect was not significant, F(1,225) = 1.72, p = .80. Among gender conforming targets, distinctiveness threat was greater for transgender (M = 3.30, SD = 1.78) than for cisgender (M = 1.45, SD = 1.08) targets, which replicates the findings of Study 2.

Finally, there was an interaction between target gender identity and target gender expression, F(1,454) = 3.89, p = .049, η _p² = .01. Simple effects revealed that there was a significant effect of gender identity on the gender expression of both male, F(1,227) = 15.38, p < .001, η _p² = .06, and female, F(1,231) = 45.23, p < .001, η _p² = .16, targets. Transgender male targets (M = 3.07, SD = 1.59) elicited greater distinctiveness threat than did cisgender male targets (M = 2.24, SD = 1.58), and transgender female targets (M = 3.29, SD = 1.64) elicited greater distinctiveness threat than did cisgender female targets (M = 1.97, SD = 1.35), supporting Hypothesis 1 that transgender targets would elicit greater distinctiveness threat than would cisgender targets. No other main effects or interactions were significant.

Moderated Moderation Analyses

As in Studies 1 and 2, we were interested in how target gender identity and target conformity affect the relationships between gender essentialist beliefs and both liking and distinctiveness threat. In the current study, we used gender essentialism as a predictor. We predicted that target gender identity and target conformity would moderate the relationship between gender essentialism and both liking and distinctiveness threat, such that transgender and nonconforming targets would be liked less and elicit greater distinctiveness threat for those scoring higher in gender essentialist beliefs. We conducted two separate moderated moderation analyses using Model 3 in PROCESS (Hayes 2013): one for liking and the other for distinctiveness threat. Gender essentialism was centered. Target conformity was coded 0 = conforming and 1 = nonconforming. Gender identity was coded such that 0 = cisgender and 1 = transgender.

Correlations among the variables in the regression model were small (see Table 2). Distinctiveness threat was significantly positively associated with target conformity and target gender identity (such that distinctiveness threat was greater for nonconforming targets and transgender targets) and with gender essentialist beliefs. Liking was significantly negatively associated with gender essentialist beliefs and distinctiveness threat, as well as was lower among targets with male gender expression.

Table 2 Correlations among variables in the regression models, Study 3

Liking

The overall moderated moderation analysis testing the moderation of gender essentialism and liking by target gender identity and target gender conformity was not significant (R² = .03, MSE = 2.05), F(7,454) = 2.01, p = .05). Target gender identity and target gender conformity were also not significant predictors of liking. The relationship between gender essentialism and liking was significantly moderated by target gender identity (β = −.14, SE = .08, t = −1.74, p = .08), such that liking decreased for transgender targets for those scoring higher in gender essentialism beliefs, which supported Hypothesis 4a (i.e., transgender targets would be liked less by those scoring higher in gender essentialism). Target gender conformity did not moderate the relationship between gender essentialism and liking (β = −.12, SE = .09, t = −1.39, p = .16), which failed to support Hypothesis 4b (i.e., nonconforming targets would be liked less by those scoring higher in gender essentialism). The test of moderated moderation was not significant (β = .13, SE = .12, t = 1.06, p = .29).

Distinctiveness Threat

The moderated moderation testing the moderation of gender essentialism and distinctiveness threat by target gender identity and target gender conformity was significant (R² = .30, MSE = 1.91), F(7,454) = 27.27, p < .001). Both target gender identity (β = 1.74, SE = .18, t = 9.61, p < .001) and target gender conformity (β = 1.34, SE = .18, t = 7.39, p < .001) were significant predictors of distinctiveness threat, such that transgender and nonconforming targets elicited more distinctiveness threat, as seen in the ANOVAs. The relationship between gender essentialism and distinctiveness threat was significantly moderated by both target gender identity (β = .25, SE = .08, t = 3.13, p = .002) and by target gender conformity (β = .19, SE = .08, t = 2.25, p = .02). As gender essentialism beliefs increased, distinctiveness threat increased for transgender targets and for nonconforming targets, supporting Hypothesis 4a (i.e., transgender targets would elicit greater distinctiveness threat for those scoring higher in gender essentialism) and Hypothesis 4b (i.e., nonconforming targets would elicit greater distinctiveness threat for those scoring higher in gender essentialism).

The overall moderated moderation analysis predicting distinctiveness threat was also significant (β = −.30, SE = .12, t = −2.59, p = .01) and increased the amount of explained variance (∆R² = .01), F(1, 454) = 6.69, p = .01 (see Fig. 3). To probe this model, we looked at the simple moderation of target gender identity within each of the conformity conditions. For conforming targets, the simple moderated model was significant (β = .25, SE = .08, t = 3.13, p < .001; see Fig. 3a), but it was not significant for nonconforming targets (β = −.05, SE = .08, t = −.61, p = .54; see Fig. 3b). Simple slopes analyses showed that higher gender essentialism was associated with higher distinctiveness threat for conforming (β = .31, SE = .05, t = 5.91, p < .001) and nonconforming (β = .20, SE = .06, t = 3.21, p < .001) transgender as well as nonconforming cisgender (β = .25, SE = .06, t = 4.36, p < .001) targets. However, gender essentialism and liking were unrelated for conforming, cisgender targets (β = .06, SE = .06, t = 1.06, p = .29; see Fig. 3).

Fig. 3

Moderation effect of target gender identity and target gender conformity on the relationship between gender essentialism and ratings of distinctiveness threat, Study 3. Black lines (both solid and dashed) indicate a significant negative relationship between prejudice and liking; the grey, dashed line indicates a nonsignificant effect for the conforming, cisgender target

Discussion

As in Studies 1 and 2, we found that distinctiveness threat was greater for transgender and nonconforming targets, regardless of gender expression. In comparing male and female gender expression directly, we found that female targets were liked less than male targets, regardless of gender identity or expression. We also found that liking was lower for transgender female targets than for cisgender female targets, but there was no effect of gender identity for male targets. There was also an interaction between gender identity and gender conformity, such that both transgender male and transgender female gender-conforming targets elicited greater distinctiveness threat than gender-conforming cisgender targets. Higher gender essentialism beliefs were associated with lower liking for transgender targets and with higher distinctiveness threat for conforming transgender targets and for nonconforming targets of both gender identities. This pattern mirrors the moderations of liking effects found in Studies 1 and 2.

Overall, the present results suggest that transgender women and transgender men are not perceived significantly differently, although transgender women seem to be liked less than transgender men. Of greater interest, Study 3 confirms that for both men and women targets, gender-conforming transgender individuals elicited greater distinctiveness threat than did gender-conforming cisgender individuals. Regardless of gender expression, it seems that “passing” transgender individuals may be threatening to binary gender boundaries. Furthermore, as we predicted, higher gender essentialist beliefs were associated with greater distinctiveness threat for conforming transgender targets. It is possible, then, that the reason “passing” transgender individuals are seen as more threatening to binary gender distinctions is because they challenge beliefs about gender essentialism and the gender binary.

General Discussion

Across all three studies, we consistently found that transgender and gender nonconforming targets elicited greater distinctiveness threat and that conforming transgender targets elicited greater distinctiveness threat than did conforming cisgender targets. These findings are consistent with prior research, which has shown that gender nonconformity is rated negatively for cisgender targets (Blashill and Powlishta 2009; Heilman and Okimoto 2007; Kapoor et al. 2010) and for nonconforming transgender targets, at least in facial appearance (Gerhardstein and Anderson 2010). Although not previously tested experimentally, our findings that gender identity-conforming transgender targets elicited greater distinctiveness threat is consistent with theoretical predictions. Because transgender individuals are perceived as transgressing binary gender roles, transgender individuals who cannot be easily detected as gender “violators” (i.e., are “passing”) may be especially threatening to binary gender distinctiveness. Our research is among the first to demonstrate that gender conformity does not necessarily affect perceptions of transgender targets the same way as it does cisgender targets. Although a large body of research has been devoted to understanding the role of gender conformity in perceptions of cisgender (including cisgender LGB) targets, and a growing body of work has investigated anti-transgender prejudice, to date, no other known research has compared the effects of gender conformity on evaluations of cisgender versus transgender targets.

Additionally, in Studies 1 and 2, target gender identity moderated the relationships between anti-transgender prejudice and traditional gender role beliefs and liking, such that higher anti-transgender prejudice and traditional gender role beliefs were associated with lower liking of transgender targets. In Studies 2 and 3, target gender identity moderated the relationships between both anti-transgender prejudice and gender essentialism with distinctiveness threat, such that higher anti-transgender prejudice and gender essentialism were associated with greater distinctiveness threat for transgender targets. We did not find any differences in attitudes toward transgender men versus transgender women; however, in Study 3, we found that nonconforming cisgender men elicited more distinctiveness threat than did conforming cisgender men, but found no differences based on gender conformity for cisgender women. This finding supports past research, which has consistently shown that male gender role transgressors are perceived more negatively than are female gender role transgressors (McCreary 1994; Sirin et al. 2004).

Overall, we found that attitudes tend to be more negative toward non-normative targets. As has been shown previously in research on anti-gay prejudice (Blashill and Powlishta 2009), our results suggest that negative attitudes toward gender role nonconformity and gender identity are largely separate. Across all three studies, main effects were found for gender conformity and gender identity, such that nonconforming targets and transgender targets were liked less and seen as more threatening. When we did find interactions between gender conformity and gender identity, the results did not indicate that negative transgender attitudes were related to gender nonconformity; indeed, the opposite was true—gender conforming transgender targets were most threatening to binary gender boundaries. Gender-conforming transgender targets elicited greater distinctiveness threat than conforming cisgender targets (Studies 2 and 3), suggesting that it is transgender men and women who are able to “pass” as cisgender who are perceived as threatening to the distinctions between binary genders. This is consistent with social identity theory, which suggests that people devalue individuals perceived as transgressing distinctive group boundaries as a means of maintaining those meaningful boundaries (Jetten et al. 1996, 1997; Van Knippenberg and Ellemers 1990) and with prior research that demonstrated that anti-transgender attitudes may stem from the need to maintain strict binary gender boundaries (Glotfelter and Anderson 2017). Additionally, this effect was bolstered for participants with stronger beliefs that gender differences are biological in nature, suggesting that at least part of the reason transgender individuals who conform to the gender roles associated with their gender expression are more threatening is because they also challenge the belief that gender is biological and immutable. By “passing” so well as to be undetectable as a transgender person, gender-conforming targets may cast doubt onto what defines gender, which threatens beliefs about gender essentialism and the gender binary.

Limitations and Future Directions

These studies were conducted to address experimental gaps in the literature on anti-transgender prejudice and how gender conformity interacts with gender identity in predicting attitudes toward transgender individuals. However, these studies are not without their own limitations. One potential limitation is that some of the vignettes were not completely balanced (e.g., stating that Jamie “acts masculine/feminine” in the male target vignettes but not in the female target vignettes). It is possible that these minor differences could change the way participants reacted to each vignette, influencing group differences in ratings of liking or distinctiveness threat. These vignettes, however, were based on validated measures created and tested by Horn (2007), with little modification made by the authors. Many of the minor inconsistencies between the vignettes are present in Horn’s original vignettes. The modifications we made were to create parallel gender-conforming male and female vignettes based on Horn’s gender nonconforming man and woman vignettes. Our modified vignettes match well with the content of Horn’s originals, at least within the target gender categories.

Additionally, we assessed attitudes toward transgender targets on only two dimensions, liking and distinctiveness threat. These outcomes were the ones of interest for our research questions, and liking is frequently used as a measure of attitudes toward specific outgroup targets in prejudice research. However, future research might benefit from investigating other attitudes toward transgender targets, such as other types of threat (e.g., danger, as seen in bathroom laws), stereotype content (e.g., mentally ill, abnormal, assumed to be gay and wanting sex reassignment surgery, unattractive; see Gazzola and Morrison 2014 for a preliminary study) or emotional reactions (e.g., disgust, discomfort; Nagoshi et al. 2008).

It should also be noted that gender identity and sexual orientation are often conflated, which could possibly affect our results. Anti-transgender prejudice and anti-gay prejudice are often correlated (Hill and Willoughby 2005; Nagoshi et al. 2008; Norton and Herek 2013; Tebbe and Moradi 2012; Tebbe et al. 2014), and according to the sexual orientation model (Herek 1984, 1994), feminine men are presumed to be gay (although masculine women are less likely to be perceived as lesbian). It is possible that participants in our study may have assumed that the transgender or nonconforming targets were gay, which led to less liking because of anti-gay prejudice. However, anti-gay prejudice is not the focus of the current studies; there are a variety of additional threats cisgender people may perceive from transgender people as the outgroup (see Branscombe et al. 1999; Cottrell and Neuberg 2005 for discussions of intergroup threat), but the current studies focus only on distinctiveness threats based on gender identity and gender conformity.

Additionally, as with most self-report survey research, there is the potential for socially desirable responding and participant bias. We also used convenience sampling in our studies, which may have affected the representativeness of our samples to the general U.S. population. Studies 1 and 2 were conducted with student samples from a private Jesuit university, which means the sample was predominantly White, wealthy, and religious. Study 3 used mTurk workers and our sample had a similar racial/ethnic breakdown as the entire country, although it was a little more liberal and less religious. It is possible that samples more representative of the U.S. population would alter our findings; however, on some key demographics known to influence anti-transgender attitudes (e.g., religiosity, conservatism), we found similar results in both the religious student samples and less religious mTurk samples. Additionally, the expression of anti-transgender attitudes remains acceptable to many people in the United States (see James et al. 2016 for anti-transgender discrimination rates; see Human Rights Campaign 2016; National Center for Transgender Equality 2015 for information on anti-transgender legislation), so it is less likely that participants felt pressure to respond in a socially desirable manner in these studies versus studies about less acceptable prejudices, such as racial attitudes. The results of our studies should also be considered within this cultural context; attitudes about gender role conformity, gender role norms, and transgender individuals likely vary between nations and cultural groups. It is possible that our results would not be replicable in other nations or cultures.

Practice Implications

Our results add confirmatory evidence that gender nonconforming and transgender individuals are viewed more negatively and that transgender individuals who conform to the societal gender roles for their gender identity (i.e., are “passing”) are viewed as the most threatening, likely because they are perceived as transgressing distinct binary gender boundaries and beliefs about gender essentialism. Thus, it is hypothetically possible to target gender essentialist beliefs using bias-reduction or counterinformational interventions aimed at reducing beliefs that there are only two binary genders and that binary gender categories are natural and immutable. Recent research has suggested that mere exposure to both information about transgender individuals and images of nonconforming and conforming transgender individuals can reduce anti-transgender prejudice (Flores et al. 2018). However, the long-term effectiveness of such interventions may be poor, if they are effective at all, especially for those with more traditional beliefs about gender. It is unclear whether gender essentialist beliefs can be altered through bias-reduction training or counter-information interventions (e.g., intersexuality, which demonstrates that even biological sex is not binary; Yoder 2012) because these beliefs are likely seen as scientific and evidenced in the natural world, and thus may be too deep-rooted to be influenced by counter-information.

Conclusion

In three studies, we investigated the role that gender conformity plays in attitudes toward transgender targets (compared to cisgender targets). In all three studies, attitudes were more negative toward nonconforming targets (regardless of gender identity) and transgender targets (regardless of gender conformity). Additionally, we found evidence that gender conformity does interact with gender identity such that gender-conforming transgender men and gender-conforming transgender women are seen as threatening to distinct gender boundaries. These effects were also augmented for those scoring higher on anti-transgender prejudice, traditional gender role beliefs, and gender essentialism. Together, these studies suggest that perceptions of transgender people transgressing or blurring distinct binary gender boundaries are threatening to gender essentialist beliefs about gender and that this is especially true for gender expression-conforming transgender individuals (i.e., those who are “passing”). Thus, anti-transgender prejudice may stem from a need to devalue and punish transgender people as a means of maintaining distinctive binary gender boundaries. Prejudice-reduction strategies and interventions should consider targeting gender essentialist beliefs in an attempt to reduce the perceptions people have about transgender people as unnatural and threatening to binary gender categories.