Development of the Gay Male Heterophobia Scale

Abstract

The current study presents the initial validation of a new 20-item measure of heterophobia, a construct frequently discussed but inadequately researched in the literature on gay men’s lives. This process included initial focus groups, exploratory and confirmatory national samples of gay men. The sample included measures of homophobia, rejection sensitivity, discrimination history, gay identity development, and social desirability to address convergent and discriminant validity. Results of an exploratory factor analysis yielded three meaningful factors: Disconnectedness, Expected Rejection, and Unease/ Avoidance. These factors had significant positive associations with constructs such as rejection sensitivity, homonegativity, and previous experiences of gay-related discrimination, but were only weakly associated with gay identity development. Heterophobia factors had insignificant or negative correlations with social desirability. This factor structure was supported by the results of confirmatory factor analyses. Implications for future research and clinical practice, along with limitations, are provided.

Gay men have commonly reported elevated rates of isolation and loneliness, particularly among adolescents (Mudrey & Medina-Adams, 2006; Radkowsky & Siegel, 1997) and older individuals (Addis, Davies, Greene, MacBride-Stewart, & Shepherd, 2009). Further studies have suggested that gay adolescents are at increased risk for social isolation in high school settings (Mudrey & Medina-Adams, 2006), experiencing lower social status and less social support than their heterosexual peers (Hatzenbuehler, McLaughlin, & Xuan, 2012). Moreover, young gay males “may be particularly vulnerable to isolation . . . in their interpersonal relationships” (Hatzenbuehler et al., 2012, p. 1189). Gay male undergraduates are more likely to be socially anxious and anticipate negative evaluation than heterosexual undergraduates, especially in situations in which stereotypically masculine behavior is valued (e.g., athletic competition and fraternity events; Pachankis, Goldfried, & Ramrattan, 2008). Moreover, gay men tend to perceive overall lower social support than lesbians, which is associated with lower resilience and greater impact of negative discrimination (Potoczniak, Aldea, & DeBlaere, 2007).

Although homonegativity and sexual stigma are a clear barrier in gay men’s abilities to form friendships with straight men, a second possible explanation is that gay men may avoid or withdraw from friendships with heterosexual men. This may in part be due to their own fears or concerns about sexual stigma or rejection from this group. Haldeman (2006) has used the term gay male heterophobia to describe this phenomenon. In his formulation, gay male heterophobia comprises a constellation of negative feelings and behaviors in gay men regarding straight men, particularly avoidance and anxiety. Haldeman (2006) further noted that heterophobia often arises from past experiences of prejudice and may serve an adaptive function. In a clinical case study, Haldeman (2006) outlined a course of treatment for heterophobia that included discussion of its origin in the client’s life, analysis of its current usefulness/necessity, and desensitization and coping strategies. Haldeman further noted that a crucial factor in the client’s treatment was the “corrective experience” of a warm relationship with a male therapist.

Previous research has suggested that attitudes similar to heterophobia, such as gay-related rejection sensitivity and internalized homonegativity, mediate the relationship between discrimination experiences and mental health problems (Feinstein, Goldfried, & Davila, 2012; Herek, Gillis, & Cogan, 2009; Pachankis et al., 2008; Rivers, 2004). Rivers (2004), for example, showed that internalized homonegativity is linked with negative societal messages about nonheterosexual identities, potentially provoking feelings of worthlessness and shame. Further research has suggested that sexual minorities with more discrimination experiences are more likely to negatively judge their sexual identities (internalized homonegativity) and to anticipate continued future discrimination (rejection sensitivity; Feinstein et al., 2012). Therefore, we propose that gay male heterophobia may comprise a set of negative cognitions about oneself and others in addition to the affective and behavioral dimensions Haldeman (2006) noted. Given the multidimensionality of this construct and potential links to important outcomes such as internalized homonegativity, a valid measure of heterophobia might be useful in a variety of clinical and research applications.

Development

In conceptualizing this construct, we initially relied on the work of Herek et al. (2009). These authors proposed that sexual attitudes (such as internalized sexual stigma) can have cognitive, affective, and behavioral dimensions. Callender (2015) further extended the cognitive–affective–behavioral framework to his study of antigay bias in the United States. Callender posited that this framework helpfully describes attitudes about sexual minority groups, encompassing “stereotypes” (cognitive), “prejudice” (affective), and “discrimination” (behavioral). With some reframing, this same three-dimensional structure can provide an initial basis for understanding heterophobia. In this case, cognitions may comprise stereotypes about straight men, whether negative (“Straight men are aggressive”) or neutral to positive (“Straight men are confident”). Cognitions may also be represented by pessimistic expectations or assumptions about relationships with straight men, for example, “That guy would never be my friend.” In the affective domain, heterophobia is best conceptualized as a set of aversive, anxious, or fearful emotions regarding interactions with straight men. Finally, the behavioral dimension may include social avoidance or other shielding behaviors such as hiding or downplaying one’s sexual identity. Here it is important that avoidance of straight men not be conflated with “discrimination” as in Callender’s model, but rather understood as a protective strategy that may or may not be helpful given the situation. This three-factor framework was used to guide our development of potential survey items, which were then tested via standard instrument development processes.

As evidence of convergent validity, we hypothesized that heterophobia would share key similarities with a number of other constructs. These scales included Rejection Sensitivity (RS), Heterosexist Harassment, Rejection, and Discrimination (HHRD), Internalized Homonegativity (Modern Homonegativity Scale, or MHS), and LGB Group Identity (Lesbian, Gay, and Bisexual Identity Measure, or LGBIM). Our rationale for including each scale was based on its conceptual or theoretical connection to heterophobia as defined earlier.

For example, we expected to see strong positive correlation between previous discrimination and heterophobia, anticipating that heterophobic attitudes may develop as a defense against adverse social conditions. Regarding rejection sensitivity, we anticipated that one’s sensitivity to rejection from straight men would also manifest in a measure of general sensitivity, and vice versa. Further, we expected that heterophobia would correlate positively with homonegative attitudes. For this sample (gay-identified adult men), we conceptualized high scores on the MHS scale as indicators of internalized homonegativity—that is, feelings of shame and discomfort with one’s gay identity, or rejection of one’s own value as a gay man. Though little is known about how internalized homonegativity might impact gay men’s relationships with straight men, we speculated that highly homonegative gay men may have difficulty in such relationships due to fear of discovery. In theory, gay men who are less comfortable in their sexual identity may experience more unease or expectations of rejection from others, given their own ambivalence about being gay. Finally, we hypothesized that identification with the gay community (captured in the LGBIM) would correlate positively with heterophobia subscales, hypothesizing that men who were more identified with a primarily gay social circle may experience more discomfort with straight men. As a measure of discriminant validity, we measured social desirability using the Social Desirability Scale–Short Form (SDS), expecting that social desirability would have no correlation with heterophobia scores.

Preliminary Focus Group Interviews, Item Development, and Pilot Testing

To assist with generating a promising item pool, we conducted focus group (before item construction) and pilot interviews (after item construction). The focus groups had three related goals. The first was to see if participants identified with the construct of heterophobia, that is, whether the concept made sense and had relevance to their experiences and worldviews. The second purpose was to gain more detailed data about the specific language that gay men use when talking about experiences with straight men that might aid in the development of our item pool. Third, we sought feedback from gay men on our original conceptualization for the ways in which heterophobia could be expressed and relatedly measured. The purpose of the pilot test interviews was to get more specific feedback on our items before the national survey.

Our focus group protocol included items in three key domains, including behavioral (“How often do you seek out friendships with straight men?”), cognitive (“What assumptions do you have about how straight men will interact with you?”), and affective (“How do you feel when you walk into a group composed primarily of straight men?”). Although questions were offered in a semistructured format and varied slightly from group to group, each session covered these three domains and addressed consistent themes (e.g., family relationships, work dynamics, and changes in relationships with straight men over time). In certain cases, responses from group participants were developed into specific scale items. For example, a number of participants discussed experiencing anxiety around straight men in locker rooms, bathrooms, and gyms, often attributed to lowered physical boundaries in these settings (e.g., changing or showering near one another). Although not all participants agreed that such situations evoked anxiety for them, the theme was prominent enough across focus groups to warrant inclusion in our initial items (“The idea of going to a predominantly straight gym makes me anxious”) and has been retained in the final scale.

Participants

For the Focus Group Interviews, 16 gay men were recruited from the adult population in a large southwestern city. Recruitment occurred primarily through the use of convenience sampling and e-mail requests to adult gay men within the local community. As an incentive to participate, participants were provided a free meal during the focus group session.

Procedure

Each focus group session was digitally recorded and transcribed, and a team of three researchers (one professor of counseling psychology, identifying as a straight male, and two graduate students in counseling psychology, both identifying as gay males) reviewed the transcripts. From our analysis, we found that some participants tended to favor the word “discomfort” or “unease” rather than “phobia” to describe their negative experiences with straight men. At the same time, other participants described feeling “fearful” in interactions with straight men and noted debilitating experiences of panic and dread. Given this complicated picture, we carefully weighed whether or not the term heterophobia should be retained as the label for this construct. After consultation and deliberation, we decided to retain the term as a means of capturing more intense manifestations of this construct described by certain group members, with its roots in the Greek suffix simply suggesting “fear.” As discussed elsewhere in the study, the term is not intended to convey pathology or signify a diagnosis, and should be applied carefully when working with this vulnerable population. In particular, the term should not be confused with homophobia, which captures a more pernicious set of attitudes held by a powerful majority population and tied to historical oppression. Rather, we conceptualize heterophobia as an understandable and even natural response to previous experiences of discrimination.

Initial Item Development

We hypothesized three underlying factors of heterophobia: Negative Cognitions, Negative Emotions, and Avoidant Behaviors related to straight men. This preliminary conceptualization was based on common notions of components underlying one’s attitude toward a range of different topics and populations. This team generated an initial list of 40 items, with each researcher contributing and refining items. Although most of these 40 items stemmed directly from the focus group themes, a number of supplementary items were added to diversify the scale content.

Each item was written in a 7-point Likert-scale format, including anchor categories of strongly disagree (1) and strongly agree (7), with 4 representing neutral, consistent with recommendations from DeVellis (2003). To encourage participants’ active reading of the items, the questionnaire was composed of an approximately equal number of positively and negatively worded statements. Further, certain items were written to express moderate to strong opinions to create more differentiation between scores (DeVellis, 2003, p. 79).

As mentioned earlier, much of the item content reflected themes emerging from focus groups. For example, focus group members voiced a range of perspectives regarding their sense of relatedness with straight men, with several members expressing the belief that straight men would not understand them. This theme was represented in the item “I think most straight men could easily relate to me,” designed as a reverse-scored item on the final Heterophobia Scale.

Pilot Testing

After we developed the initial items, we then conducted a pilot test of 11 participants using the same inclusion criteria. Eight participants identified as White, two as Latino, and one as Asian. All participants had received at least some college education, two had completed PhDs in counseling psychology, and one was a current graduate student in counseling psychology. With this part of the process we aimed to gain feedback regarding items’ clarity, brevity, and flow, and to ensure the content validity of the items, that is, to confirm their relevance to the targeted construct. Electronic copies of the initial 45 items were distributed to the 11 pilot test participants, who were asked to evaluate each item on the above criteria. Several items flagged as problematic by multiple participants were revised or removed from the questionnaire, resulting in an initial Heterophobia Scale composed of 40 vetted items.

In general, pilot test participants confirmed that scale items appeared relevant to the construct of heterophobia. However, several participants suggested that items be rephrased so that they could apply more generally to all gay men. It was also suggested that the scale be clearly defined as measuring “male-specific heterophobia” given that both the subjects and objects of this construct are men. As the primary focus of the current article is on measure development, focus group and pilot testing data are not included. However, the raw data are available upon request from the Markus M. Provence.

Study 1: National Survey and Convergent/Discriminant Validity

Following pilot testing, a large nationwide sample was recruited via a combination of methods, including e-mail requests, online advertisements, and compensated subject pools. Specifically, participants were recruited through Amazon’s Mechanical Turk (Mturk), an emerging online service connecting researchers and paid survey participants. In a recent study, Bartneck, Duenser, Moltchanova, and Zawieska (2015) identified Mturk as a valid and efficient recruitment method for social science research, finding no significant difference in survey responses between Mturk participants and those recruited through more traditional online methods. However, Fleischer, Mead, and Huang (2015) argued that participants using Mturk may answer items inattentively, and urged caution for researchers using this method. Data collected using this recruitment method therefore carries some limits to validity, as mentioned further in our Discussion section.

As a supplement to Mturk, targeted Facebook advertisements were used to promote the study to men who identified romantic interest in other men. As an incentive to participation, Facebook and convenience sampling recruits were entered into a raffle to earn three $100 Amazon gift certificates. Participants entered this raffle by sending an e-mail to a specially designated Gmail account, which was kept separate from the anonymously coded questionnaires. Three raffle winners were then chosen via a random number generator.

Within both recruitment methods, the study was advertised generically as an attitudinal survey for gay men. Interested parties were provided with the following study description before taking the survey: “The purpose of this study is to learn about gay men’s attitudes and experiences. Specifically, we are interested in understanding gay men’s relationships with heterosexual men.” The term heterophobia was avoided in recruitment and survey materials.

Approximately 300 participants were recruited using Mturk, and ~150 were recruited through Facebook and e-mail networking. Due to constraints in the initial design of the study, data from both sets of participants were combined into a single sample. We did not track differences in results between the two samples.

Participants

A total of 451 participants participated and composed the pool used in the exploratory analysis. After inspecting the data, 13 cases were found to be incomplete and excluded, leaving a final total of 438 participants. Of these, 59.4% identified as White, 20.5% as Asian, 8.4% as Hispanic or Latino, 5.9% as Black or African American, 3% as biracial or multiracial, 1.8% as American Indian or Alaska Native, and .5% as other. Participants had a mean age of 31.75 (SD = 9.19; range = 18–80). Participants were asked to identify their highest educational level completed; 32% completed high school, 44.3% received at least a bachelor’s degree, and 23.5% obtained a graduate degree. Finally, participants represented a range of geographic locations, with Texas (23.3%), New York (8%), California and Florida (5.9% each) containing the highest percentages of participants. An additional 7.5% of participants identified as currently living outside of the United States.

Measures

Heterosexist harassment, rejection, and discrimination.

The HHRD Scale measures the extent to which an individual has experienced discrimination related to sexual minority status (Szymanski, 2006). The scale has been adapted for use with lesbians and gay men (Feinstein et al., 2012). Sample items from this scale include “How many times have you been treated unfairly by your family because you are gay?” and “How many times have you been verbally insulted because you are gay?” Each of the 14 items is rated on a 6-point scale from 1 (never happened to you) to 6 (happened often, e.g., over 70% of the time). Overall scores for this scale are generated by averaging the score for each item, creating a final score range of 1–6. Higher mean scores signify a higher occurrence of heterosexist harassment, rejection, and discrimination during the past year. We observed an internal consistency (Cronbach’s α) of .96 for this scale.

Experiences of rejection.

The gay-related RS Scale assesses an individual’s sensitivity to social rejection based on his or her sexual orientation (Pachankis et al., 2008). The scale consists of 14 items that describe potential social rejection scenarios. A sample scenario is as follows:

You and your partner are on a road trip and decide to check into a hotel in a rural town. The sign out front says there are vacancies. The two of you go inside, and the woman at the front desk says there are no rooms left.

After reading the items, a participant then evaluates the scenario on two levels: First, how anxious/concerned he would be if this scenario happened to him, and second, how likely it is that the event occurred because of his sexual orientation (Pachankis et al., 2008). Each item is rated on a 6-point scale from 1 (very unconcerned/very unlikely) to 6 (very concerned/very likely). Higher RS scores signify a higher sensitivity to gay-related rejection. We observed an internal consistency (Cronbach’s ) of .947 for this scale.

LGB identity.

The LGBIM assesses an individual’s level of identification with his or her particular ethnic group, as well as his or her level of acceptance/rejection of outside groups (Phinney, 1992). This scale has been adapted for use with LGB group identification (Mohr & Kendra, 2011). Each of the 20 items is rated on a 4-point Likert scale from 1 (strongly disagree) to 4 (strongly agree), with 14 items measuring same-group orientation (e.g., “I have a strong sense of belonging to the LGB community”) and six items measuring other-group orientation (e.g., “I often spend time with straight people”). The overall scale score is derived by obtaining a mean for the responses to each item. Higher LGBIM scores signify a higher identification with the LGB community. We observed an internal consistency (Cronbach’s α) of .78 for this scale.

Internalized homonegativity.

The MHS is designed to measure an individual’s negative attitudes toward sexual minorities (Morrison & Morrison, 2003). The MHS is sometimes used as an alternative to the Hudson and Ricketts’s (1980) Index of Homophobia because it captures a broader range of homonegative attitudes, uses a more modern vernacular, and contains 10 rather than 20 items (Rye & Meaney, 2010). Sample items include “Gay men have become far too confrontational in their demand for equal rights” and “The notion of universities providing students with undergraduate degrees in Gay and Lesbian Studies is ridiculous.” Each of the 10 items is rated on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree); overall scale scores are represented by the mean of the items. Higher MHS scores signify higher levels of homonegativity. We observed an internal consistency (Cronbach’s ) of .91 for this scale.

Social desirability.

The SDS is designed to measure the extent to which a person presents him- or herself positively on self-report surveys (Strahan & Gerbasi, 1972). The short form was adapted from the original Social Desirability Scale (Crowne & Marlowe, 1960) to facilitate shorter test-taking time. Each of the 10 short-form items is rated true or false, and half are reverse scored. Sample items include “You are always willing to admit it when you make a mistake” and “At times you have really insisted on having things your own way.” Higher SDS scores signify stronger desire to present oneself in a positive light. We observed an internal consistency (Cronbach’s α) of .57 for this scale. Conclusions about discriminant validity should therefore be interpreted with some caution.

Procedure

Participants read and acknowledged an informed consent document outlining potential risks and benefits of participation. They were then asked to complete a Demographic Questionnaire, which included the aforementioned questions along with several questions addressing relationship history, childhood rejection experiences, and level of openness about their sexuality. Participants then answered the question, “Do you self-identify as a gay male?” Any person who selected “no” was prevented from continuing with the survey. Following demographics, participants completed the 40-item, pilot-tested Heterophobia Scale and the other measures.

Factor analysis

Following data collection, Bartlett’s test of sphericity was used to provide evidence for sampling adequacy. For this sample, the chi-square result for Bartlett’s test was 9,923.98, suggesting that sampling adequacy had been reached and principal components analysis was an appropriate statistical method for this study.

SPSS was then used to conduct principal components analysis using a Promax rotation for the complete data set (N = 438) of the 40-item Heterophobia Scale. A number of criteria were used to select an appropriate number of underlying factors for this scale, using an iterative process of testing various factor solutions. First, using the “eigenvalue-greater-than-1” rule, we identified six potential components above this cutoff. An inspection of the scree plot (Cattell, 1966), however, suggested that a three- or four-factor solution may be more appropriate, as the fourth, fifth and sixth factors accounted for relatively little additional variance after the first three. Further, results of a parallel analysis (Hayton, Allen, & Scarpello, 2004) also supported three- or four-factor solutions.

A four-factor solution was tested, including analysis of individual item loadings on each factor. Item loadings above .5 were initially favored, with three factors comprising the vast majority of these loadings. Factors one, two, and three were associated with high loadings (>.5) on 38 of 40 items, whereas factor four contained only two highly loaded items. Therefore, a three-factor solution (as initially hypothesized) was tested; in this case, high item loadings were now distributed representatively between the three factors. The full list of 40 original items with factor loadings is presented in the online supplemental materials.

Grouped by their associated factor, items suggested interpretable thematic categories. To determine these categories, individual items within each factor were inspected and labeled with a summary word or phrase. These phrases were then analyzed and compared until common themes emerged. The themes were then cross-referenced with relevant theory and research on heterophobia, and various factor labels were considered.

Interestingly, the ultimate categories did not correspond to the hypothesized factors of negative cognitions, negative emotions, and avoidant behaviors. Instead, the items cohered into unexpected, but nonetheless meaningful themes. These included Unease/Avoidance, Disconnectedness, and Expected Rejection (described in detail later in the text). Correlations between each factor were within acceptable ranges. The correlation between Unease/Avoidance and Expected Rejection fell in the relatively high range (r = .62) but was low enough to suggest that these factors are still meaningfully distinct. This is corroborated by an inspection of the items themselves, which seemed conceptually related but also differentiated enough to warrant separate subscales. The full list of interfactor correlations is presented in Table 1.

Table 1.

Intercorrelation Matrix Including Subscales and Additional Measures

Scale	1	2	3	4	5	6	7	M	SD
1. U/A								3.44	1.55
2. DIS	.28**							3.23	1.23
3. ER	.62**	.29**						4.65	1.25
4. HHRD	.67**	− .06	.39**					2.56	1.20
5. RS	.51**	.06	.50**	.49**				15.07	7.69
6. LGBIM	.17**	.09	.11*	.24**	.33**			2.65	0.38
7. MHS	.46**	−.23**	.20**	.56**	.23**	−.15**		2.41	0.96
8. SDS	−.13**	.09	0	−.11*	− .07	−.05	−.19**	1.55	0.20

Note. U/A = Unease/Avoidance; DIS = Disconnectedness; ER = Expected Rejection; HHRD = Heterosexist Harassment, Rejection, and Discrimination; RS = Rejection Sensitivity; LGBIM = Lesbian, Gay, and Bisexual Identity Measure; MHS = Modern Homonegativity Scale; SDS = Social Desirability Scale-Short Form.

^*p < .05.

^**p < .01.

After factor extraction, we used several criteria to reduce the overall item count without losing high reliability. First, items were evaluated with regard to content validity within a given factor—that is, we deleted items that did not appear to be interpretable within the factor’s conceptual theme. Further, items were examined for redundancy using the interitem correlation matrix. When two items had very high correlations with one another (e.g., items 30 and 34, r = .77), at least one of the items would be flagged as a candidate for deletion. A further criterion for selecting or deleting items was changes in certain key statistics when items were deleted, including percent of total variance explained and Cronbach’s. As a final guideline for selecting items, we examined item-to-total correlations and removed the lowest-scoring items.

A final list of 20 items was generated based on the aforementioned criteria. Factor loadings for each item ranged between .59 and .90. (See Table 2 for a list of factor loadings on the revised 20-item scale.) To gauge each scale’s internal consistency, the researchers calculated the Cronbach’s α coefficient for each factor subscale. A brief description and rationale for each factor is included in the following text.

Table 2.

Item Loadings and Factors for Revised Heterophobia Scale

Item content	F1	F2	F3
Factor 1: Unease/Avoidance
7. When I’m the only gay man in a social setting, I feel unsafe.	.799	.021	.029
9. I become uneasy making small talk with straight men.	.831	.086	− .036
10. The idea of going to a predominantly straight gym makes me anxious.	.828	− .086	.011
12. I feel tense in my interactions with most straight men.	.880	−.017	− .009
14. I believe that a straight man may behave violently toward me if he knew that I was gay.	.717	− .067	−.147
34. I tend to avoid straight men.	.862	.019	− .048
40. When using a restroom, I am careful to keep my distance from straight men.	.621	.012	.096
Factor 2: Disconnectedness
1. In social situations, I’m just as comfortable being with straight men as gay men. (R)	.101	.774	− .063
4. I feel equally free to be myself among gay men and straight men. (R)	−.054	.709	.031
13. It is easy for me to enjoy myself when spending time with straight men. (R)	.238	.734	−.120
15. Straight men share my basic values. (R)	−.152	.742	.070
17. I think most straight men could easily relate to me. (R)	− .304	.712	.323
29. I would readily join a group or club that included mostly straight men. (R)	.048	.759	−.055
31. I am equally likely to interact with gay or straight men at social gatherings. (R)	.087	.759	−.140
Factor 3: Expected Rejection
19. When interacting with a straight man, I tend to wonder whether he will accept me.	.289	−.017	.591
21. Straight men wouldn’t want to hear about my coming out experience.	.146	.010	.689
23. I think some straight men might feel uncomfortable if they know I’m gay.	.088	.018	.709
26. Straight men would judge me if they found out about my sexual orientation.	.197	.035	.683
27. Straight men would be put off by hearing the details of my sexual life.	− .221	− .097	.897
38. I don’t talk about dating with straight men.	.074	.057	.692

Note. (R) = reverse-scored item.

Factor 1: Unease/Avoidance.

The seven items in this subscale seemed to relate to negative affective states (anxiety, unease, and tension) and desire for avoidance of straight men. Conceptually, this subscale might be said to capture gay men’s more “visceral” or emotional reactions to straight men and attempts to cope with these emotions through physical separation. The mean score for this subscale is 3.43 (suggesting overall scores in the moderate range). Cronbach’s α for responses to items on the Unease/Avoidance factor was .91. (Sample item: “When I’m the only gay man in a social setting, I feel unsafe.”)

Factor 2: Disconnectedness.

The seven items in this subscale address gay men’s feelings of closeness or estrangement in relation to straight men. Themes addressed in these items included gay men’s sense of shared values with straight men and their openness to interacting with straight men socially. Items on this subscale were reverse scored, with higher scores indicating less agreement with the item (and therefore higher disconnectedness from straight men). Although it is interesting to note that all reverse-scored items loaded onto the same factor, they do appear to be thematically coherent, meriting their inclusion together. The mean score for this subscale is 3.23, suggesting that it is the least endorsed of the three factors. That is, participants in this sample were slightly less likely to identify disconnectedness to straight men than unease/avoidance or expectations of rejection. Therefore, disconnectedness seems to be a marginally less relevant concern. Cronbach’s α for responses to items on the Disconnectedness factor was .86. (Sample item: “Straight men share my basic values.”)

Factor 3: Expected Rejection.

This subscale measures the extent to which gay men anticipate straight men reacting negatively to their sexuality. These six items tend to capture negative appraisals about how straight men might respond to them when discussing sexuality or dating, or simply revealing that they are gay. Although this subscale has some conceptual similarity with Unease/Avoidance, a key difference is that the items here described cognitions about what might happen, whereas the Unease/Avoidance subscale primarily describes emotions and behaviors. The mean score for this subscale is 4.66, indicating that this subscale had the highest overall mean scores. Stated another way, respondents were overall more likely to endorse anticipating rejection from straight men than to endorse unease/avoidance or disconnectedness. Cronbach’s α for responses to items on the Expected Rejection factor was .86. (Sample item: “Straight men wouldn’t want to hear about my coming out experience.”)

The α was calculated for each factor at various points in the iterative factor analysis and item removal process. The final α calculations (.91, .86, and .86) provide evidence that each subscale demonstrated adequate levels of internal consistency and may be used independently. Finally, data were assembled into an intercorrelation matrix containing eight measures, including the three subscales and five additional measures. We calculated correlations (Table 1) based on factors rather than the overall Heterophobia Scale because each factor was judged to be relatively distinct. Individual correlations were tested for significance at p < .01.

Correlations between heterophobia factors and additional scales provide preliminary evidence for convergent and discriminant validity. The Unease/Avoidance and Expected Rejection factors were positively correlated with a number of related scales, including HHRD Scale (indicating previous experiences of gay-related discrimination), RS Scale, and MHS. Interestingly, the Disconnectedness factor did not correlate strongly with any of these scales; therefore, convergent validity is somewhat more difficult to establish for this subscale. All three Heterophobia factors had weak but statistically significant correlations (p < .01) with LGB group identification (captured by the LGBIM scale), providing tentative evidence for the discriminant validity of each factor. Finally, the SDS had insignificant or negative correlations with the Heterophobia factors, suggesting that participants responded honestly without regard for others’ perceptions.

In addition to the supplemental scales, participants were asked to complete a Demographic Questionnaire including basic personal information and two questions about past experience, rated on a 1–7 Likert scale: (a) “I have been satisfied with my dating/relationship experiences with other men” and (b) “Growing up, I experienced rejection from straight men.” Finally, respondents were asked to describe their level of openness about being gay (“Outness”), also on a 1–7 scale, from 1 (completely hidden from others) to 7 (completely open with others).

Study 2: Confirmatory Factor Analysis and Test–Retest Analysis

Participants

Data from 283 self-identified gay men, collected from Mturk, were used in the analyses. Regarding race/ethnicity, 68% of participants identified as White/Caucasian, 9% as Hispanic/Latino, 9% as Black or African American, 8% as Asian/Asian American, 3% as Native American, 3% as multiracial, and less than 1% as Native Hawaiian. Participants averaged 31.97 years old (SD = 8.79; range = 18–72).

Measures

The 20-item Heterophobia Scale was administered to participants. Items and response options are described in Study 2.

Procedure

Data were collected on Mturk. Participants were able to view the current study opportunity from a list of opportunities available to them. The informed consent document stated that the study was about gay men’s interactions with others. Participation was restricted to individuals who were located in the United States and who had at least 95% approval on previous Mturk tasks. The survey contained validity check items (e.g., “Please check disagree”). Failure to correctly answer a validity check item resulted in termination of the survey, and data from individuals who failed validity check items were not used in analyses. Of the initial 315 participants, 32 failed a validity check item or identified as a sexual orientation other than gay on the demographics page, and were not included in any analyses.

Of the participants who completed the data collection for the confirmatory factor analysis (CFA), we invited a random subset of 120 participants to participate again to obtain test–retest data. We aimed to collect 120 participants to account for expected attrition, but lost only one participant to follow-up. Thus, the final sample used to calculate test–retest data was 119. The second wave of data collection was posted exactly 3 weeks after the first data collection, and all participants completed the second wave of data collection within 36 hr of the survey being released. Participants who completed the second wave of data collection received an additional $1.00 toward their Mturk account. Participants were matched across time points by use of user-generated personal but nonidentifiable codes (a short set of strings of data, such as “enter the last two letters of your father’s name”; Garvey Wilson et al., 2010; Yurek, Vasey, & Havens, 2008).

Confirmatory Factor Analysis

For the CFA, a total of 10 missing data points were present out of 5,660, representing a missing data rate of 0.18%. For test–retest data, a total of nine data points were missing out of 714, representing a missing data rate of 1.26%. Because of this low level of missing data, we used full information maximum likelihood estimation to conduct analyses in Mplus and available item analysis to conduct internal consistency and test–retest reliability analyses (Parent, 2013). Consistent with recommendations for assessing fit in CFA models, acceptable fit would be indicated by a confirmatory fit index (CFI) greater than .90, residual mean square error of approximation (RMSEA) with an upper bound under 0.10, and a standardized root mean residual (SRMR) under 0.10. Good fit would be indicated by a CFI over .95, an RMSEA with an upper bound under 0.06, and an SRMR with a value under 0.08 (Weston & Gore, 2006).

For the CFA, data were analyzed in Mplus Version 7.2 (Muthén & Muthén, 1998–2015) using maximum likelihood estimation with robust standard errors. Five multivariate outliers were identified using Decarlo’s implementation of Mardia’s test, but removal of these cases had no effect on the analyses and so they were retained. The thee-factor model was established as per Study 1, with each item constrained to load onto its intended latent factor, no item covariances estimated, and the three factors allowed to covary. This model fit the data within acceptable parameters, χ² (167) = 373.02, p < .001; CFI = 0.90; RMSEA = 0.07 (90% confidence interval = 0.06–0.08); SRMR = 0.07. These values fell within the acceptable ranges for CFI and RMSEA, and the good range for SRMR. Standardized factor loadings are presented in Figure 1. We conducted exploratory tests of other models (e.g., a higher order factor model and a bifactor model), but no alternative models demonstrate acceptable fit indices.

Figure 1.

Confirmatory factor analysis standardized loadings for the Heterophobia Scale.* p < .01. ** p < .001.

Test–Retest Reliability

Regarding test–retest data, mean scores on the measures were calculated, with higher scores indicating greater endorsement of the three heterophobia subscales. Test–retest reliability was obtained by assessing the correlations between scores at Time 1 and Time 2. In this sample, 3-week test–retest reliabilities for the subscales were r = .78, p < .001, for Unease/Avoidance; r = .60, p < .001, for Disconnectedness; and r = .70, p < .001, for Expected Rejection. These results suggest that the subscales measure participant attitudes within a 21-day time period with acceptable reliability.

Overall Discussion

Overall, the results of this study provided initial evidence for the reliability and validity of the Heterophobia Scale. The factor analysis results suggested three underlying components of heterophobia, described as Unease/Avoidance, Disconnectedness, and Expected Rejection. Internal consistency results for the subscales were all acceptable (i.e., between .86 and .91). Intercorrelations between the factors were all significant, showing a positive relationship between each subscale. At the same time, the correlations suggested that the subscales represented meaningfully distinct constructs. The factors did not align with our expectation of cognitive, affective, and behavioral dimensions, but nonetheless captured clear and resonant aspects of our construct.

The emergence of these specific factors provides a more nuanced lens for understanding heterophobia. Rather than understanding the construct as composed of generic affective, cognitive, and behavioral dimensions, we now better understand what heterophobia may look and feel like for gay men. Following this insight, we can begin to ask questions in a more nuanced manner about how disconnectedness, unease/avoidance, and expected rejection impact gay men’s sense of themselves in the world. For example, unease/avoidance may be particularly salient in work situations, whereas disconnectedness might relate to one’s overall feelings of isolation. Preliminary correlation data from the study hint at the richness of these constructs.

Strong positive correlations between heterophobia factors and additional scales, including Rejection Sensitivity, Heterosexist Harassment, Rejection, and Discrimination, and Homonegativity, were found, providing evidence for convergent validity. These relationships matched our expectations and suggested that heterophobia factors had only negative, insignificant, or weak correlations with gay group identification and social desirability, providing tentative evidence of discriminant validity.

A number of important implications for future research and clinical practice can be noted. For one, it is critical to better understand the relationship between heterophobia and various mental health indices. We know when compared with straight men, gay men are at elevated risk for depression and suicide and tend to engage in more risk-taking health behaviors (Hamilton & Mahalik, 2009; Lewis, 2009; Livingston et al., 2015). It may be telling to initiate research that may address whether the construct of heterophobia is an important factor in predicting what types of men might be more vulnerable to such negative mental health and behavioral pattern outcomes. Considering the possible value friendships typically have in protecting against mental health disorders (Fenaughty & Harré, 2003; Paul et al., 2002), a relationship between such variables may be plausible.

In addition, researchers may use the heterophobia subscales to investigate new questions about gay men’s lives and relationships. Included in these future research threads might be a need to test hypothesized relationships between heterophobia factors and other relevant constructs. For example, it is important to determine the extent to which unease/avoidance of straight men contributes to overall feelings of loneliness/isolation in gay men’s lives. Researchers might also test whether heterophobia factors serve as mediators between personally and vicariously experienced rejection, harassment, or violence, and key outcomes such as suicidality, depression, anxiety, stress, and life satisfaction. We would also like to see further investigation of the relationship between heterophobia factors and related constructs cited in this study, such as internalized homonegativity (Greene & Britton, 2015), heterosexual self-presentation (Parent & Moradi, 2009, 2011), and rejection sensitivity (Feinstein et al., 2012). Finally, it would be useful to test for potential moderators of the relationship between rejection experiences and heterophobia factors. Moderators to be tested might include stage of gay identity development (Cass, 1984), trauma history (Roberts, Austin, Corliss, Vandermorris, & Koenen, 2010), and adherence to masculine norms (Levant, 2011; Parent & Bradstreet, 2017). It may also be interesting to investigate relationships between heterophobia scores and other demographic factors such as income level or number of straight male friends.

We have chosen to retain each of the three factors with the belief that each subscale represents a unique and important construct of its own. For example, we might expect Disconnectedness to correlate with relationship satisfaction or feelings of emotional intimacy with straight family members and friends. This subscale may even provide a helpful method for assessing and predicting a gay male client’s rapport with a straight male therapist. It is assumed that the Unease/Avoidance subscale might correlate with behavioral measures such as willingness to venture into “straight male” environments. Further, this factor is expected to identify more “gut-level” anxious and aversive responses to straight men, which may be relevant to assessing the impact of trauma. Finally, the Expected Rejection subscale appears useful for identifying cognitive schemata that impact gay men’s willingness to connect with straight men. In clinical settings, such a measure may highlight areas for further exploration, for example, suggesting adaptive or less adaptive “automatic thoughts” in cognitive-based therapies.

Further, a number of experimental studies might be developed using validated heterophobia measures. For example, researchers could test whether one’s level of disconnectedness responds to a certain type of treatment, such as increased positive contact with straight men. In this scenario, gay men might take a pretest measuring disconnectedness, then participate in a 12-week mixed-orientation men’s therapy group, followed by a posttest using the same measure. A significant decrease in disconnectedness scores might then provide tentative evidence for the effectiveness of this treatment. Other treatments might include participation in outdoor adventure therapy with straight men or engaging in interpersonal therapy with a straight-identified individual therapist.

A number of other compelling experimental studies may be developed using these subscales. First, given the increasing popularity of “third-wave” cognitive therapy approaches, it might be interesting to test whether therapeutic approaches such as acceptance and commitment therapy (Villatte et al., 2016), mindfulness meditation (Shearer, Hunt, Chowdhury, & Nicol, 2016), and guided imagery (Jennings & Jennings, 2013) might impact unease/avoidance scores over time. From an educational perspective, researchers might test whether heterophobia factors predict academic performance in different kinds of learning environments. For example, a study could be designed to investigate whether expected rejection impacts test scores for gay students in classes with straight male teachers. Although these ideas provide some initial suggestions for future research directions, it is hoped that this measure will be used in innovative ways.

This instrument development project also has preliminary implications for clinicians. For example, counselors might also use heterophobia research to develop interventions for clients who identify relationships with straight men as problematic in their lives. (Here we would caution clinicians to view heterophobia not as a sign of pathology, but rather as a natural response to past experiences of sexual stigma from straight men.) Promising techniques might be drawn from cognitive–behavioral, interpersonal, and mindfulness-based approaches to help clients gain greater awareness of heterophobic thoughts and feelings and move closer to their personal goals. Counseling with heterosexual therapists may be especially helpful in helping heterophobic clients face their fears in a safe environment. Straight male therapists who use a nurturing and gay-affirming style may have much to offer heterophobic male clients in this regard. In developing trusting relationships with members of a previously feared group, gay men may develop “schemas of acceptance” that enhance their cross-orientation friendships and even their same-sex partnerships (Pachankis et al., 2008). Supportive contact with straight men in group counseling has been suggested as a potential “corrective experience” for gay men experiencing heterophobia (Provence, Rochlen, Chester, & Smith, 2014). Further, counselors working with gay male couples might look to heterophobia research to gain more insight into potential sources of relationship strain. Given that higher stigma consciousness is associated with poorer relationship outcomes in gay men (Doyle & Molix, 2014; Peplau & Fingerhut, 2007; van Eeden-Moorefield & Benson, 2014), couples counselors working with this population may want to address signs of expected rejection from straight men. Again, it is critical for therapists not to conceptualize heterophobia as a pathology or disorder in their clients, but to seek better understanding of its origins and consequences in a client’s life.

Despite its strengths, the current study has several key limitations. First, the sample demographics disproportionately represent educated White men under 40, and so do not capture a representative population of American gay men. Further, we did not collect income data for this sample and thus do not know how socioeconomically diverse this sample may be. As such, the results from this study are not fully generalizable to all gay men and should be interpreted with caution. In addition, we chose to combine sample data from two separate recruitment methods (Mturk and online networking via Facebook and e-mail); therefore, validity results should be interpreted carefully. The validity of the results may also have been impacted by the lack of attention checks for online survey participants in Study 1, though we did implement these checks in Study 2.

A further study limitation is that our construct is restricted to a narrow range of attitudes—those experienced by gay men in relation to straight men. Although this dynamic is undoubtedly an important one, it is by definition limited. For example, we exclude the experiences of lesbian, bisexual, and other sexual minority individuals who may also experience heterophobia. Further, the current study does not address gay men’s feelings toward heterosexual women, which may be an important corollary to male-specific heterophobia. It is hoped that future studies will broaden the current findings to include these populations and relationships.

In summary, this study provides compelling evidence for the relevance of heterophobia and its factors. It is hoped that this study will increase awareness of, and compassion for, gay men’s unique experiences among a more powerful majority of straight men. Given the strong connections between heterophobia and both childhood and adult rejection experiences, there is clearly more work to be done in the domain of gay-affirmative advocacy and social justice. Social workers, teachers, school counselors, parents, employers, religious leaders, and policymakers may seek to better understand how these painful experiences go on to shape the worldview and well-being of gay men and boys in their midst.