Introduction

Gay and bisexual men who are gender-nonconforming or display feminine characteristics have consistently been shown to experience more mental distress than gay and bisexual men who are gender-conforming (D’Augelli, Grossman, & Starks, 2006; Grossman, D’Augelli, Salter, & Hubbard, 2005; Henning-Stout, James, & Macintosh, 2000; Landolt, Bartholomew, Saffrey, Oram, & Perlman, 2004; Martin-Storey & August, 2016; Ploderl & Fartacek, 2009; Sandfort, Melendez, & Diaz, 2007; Skidmore, Linsenmeier, & Bailey, 2006). Studies have also indicated that this elevated mental distress results from higher levels of discrimination experienced by gender-nonconforming gay and bisexual men (Baams, Beek, Hille, Zevenbergen, & Bos, 2013; Martin-Storey & August, 2016; Sandfort et al., 2007; Toomey, Ryan, Diaz, Card, & Russell, 2010; Van Beusekom, Baams, Bos, Overbeek, & Sandfort, 2016).

Thus far, the relationship between gender nonconformity and mental distress has primarily been studied in samples of gay and bisexual men living in Western countries. In a study among Black men who have sex with men (MSM) in South Africa, Cook, Sandfort, Nel, and Rich (2013) found no support for the observed associations. As reported in a Letter to the Editor of this journal, gender-nonconforming men indeed suffered higher levels of discrimination relative to gender-conforming men; furthermore, discrimination was positively associated with mental distress in the total sample. However, despite these associations, gender-nonconforming MSM did not have more mental distress than gender-conforming MSM.

In a subsequent study, using a different sample of Black South African MSM, Sandfort, Bos, Knox, and Reddy (2016) also found that gender nonconformity was not associated with mental distress, despite the fact that gender-nonconforming men experienced more discrimination and that discrimination was associated with mental distress. In addition, this study found an indirect effect of gender nonconformity on depression through internalized homophobia, suggesting that, in this population, internalized homophobia overrode the effect of discrimination on mental distress.

Even though this second study contributed to a further understanding of the distinctive relationship between gender nonconformity and mental health among Black South African MSM, several questions remained unanswered. We wondered whether it mattered whether gender nonconformity was operationalized as a categorical variable instead of a continuous variable, as we did in Sandfort et al. (2016). In that study, we used a categorical variable, a preference for a feminine or masculine gender presentation, to validate the assessment of gender nonconformity as a continuous variable. We showed that MSM who preferred to present themselves to others as feminine had the highest gender nonconformity score and differed significantly from MSM who preferred to present themselves as masculine and MSM who reported no specific preference for gender expression. Men who had no preferred gender expression also scored higher on gender nonconformity than men with a masculine gender expression. Furthermore, we wondered whether the stressors and resilience factors have the same impact on mental distress across the three gender expression groups as a categorical variable. Exploring this could shed further light on the unexpected findings from Cook et al. (2013) and Sandfort et al. (2016). Using the same sample of Black South African MSM as in Sandfort et al. (2016), we explored whether there were differences in mental distress, and stressors and resilience factors between the preferred gender expression groups (feminine, masculine, and no preference) and whether gender expression moderated the associations of the stressors and resilience factors with mental distress.

Method

Participants

We used multiple recruitment strategies to recruit a heterogeneous sample of Black South African MSM based on age (MSM between 18 and 25 years of age and above 25 years) and residential status (MSM living in townships and those living in urban areas), with townships being characterized by low levels of education, high unemployment, and more poverty. Because the level of MSM community organization is low and there is no MSM commercial subculture, we held social functions for Black MSM throughout the township. Black men living in the urban area were invited to attend social events at an LGBT community center. Eligibility criteria included (1) living in the greater Pretoria metropolitan area; (2) being between 18 and 40 years old; (3) identifying as Black or African; (4) reporting to have had oral, anal, or masturbatory sex with at least one man in the preceding year, regardless of involvement with women and including men who did not self-identify as gay; and (5) being conversant in English. Participant recruitment and data collection were conducted from October to December 2008.

A total of 199 Black South African MSM were surveyed; three participants were excluded because they did not provide information on gender expression, resulting in an analytic sample of 196. The men ranged in age between 18 and 40 years, and the mean age was 26.65 (SD = 5.59). Seventy-nine percent (n = 148) of the participants lived in a township. Two-thirds (n = 119) were in an ongoing intimate relationship with a man, and the average duration of these relationships was 2.87 years (SD = 1.07). The majority of men (63.8%, n = 125) were employed. Twenty-nine percent (n = 57) of the participants had no income, and one-third (33.7%, n = 66) had a low income (i.e., less than 4501 South African Rand per month). Sixty-eight percent (n = 132) reported that they were religious.

Of the 196 men, 57 (29.1%) preferred to present themselves as feminine, 77 (39.3%) as masculine, and 62 (31.6%) did not have a specific preference (subsequently also indicated as men with a neutral gender expression). We compared these three groups on demographic characteristics (see Table 1) and found that, compared to masculine men, feminine men were younger and were more frequently involved in an ongoing intimate relationship with another man. A higher proportion of feminine men lived in a township (instead of the city center), were unemployed, and had a lower monthly income compared to masculine men and men with no preference. Men with no preference were significantly younger compared to masculine men.

Table 1 Sample description of Black South African MSM by gender expression
Full size table

Procedure

After being informed about study procedures, interviewers obtained verbal consent. Once confirmed, all participants were asked to complete a questionnaire on the spot. Privacy was maintained by having participants complete the survey in quiet, usually adjacent rooms. Interviews were administered using computer-assisted self-interviewing in order to minimize social desirability bias. Participants were compensated equal to approximately $8 USD for their time.

Measures

The survey collected information on sociodemographic characteristics, sexual orientation, gender expression, mental health, sexual minority stressors, and resilience factors. The sociodemographic characteristics measured included age, residential status (living in a township or in the city), educational attainment, income, and employment status.

Gender Expression

Gender expression was assessed by asking participants in what way they prefer to present themselves to others: feminine, masculine, or no preference. We established construct validity for this gender expression assessment by testing whether these three groups differed from each other on a scale that measured how participants perceived themselves in terms of masculinity and femininity. This masculinity/femininity (M/F) scale, adapted from Storms (1979), consisted of two items (“Do you see yourself as more masculine or more feminine than most other men?” and “Do you think other people see you as more masculine or more feminine than most other men?”; 1 = much more masculine – 5 = much more feminine; (Cronbach’s alpha = .85). Men who preferred to present themselves to others as feminine had the highest M/F score (M = 4.52, SD = 0.66) and differed significantly from men who preferred to present themselves as masculine (M = 1.99, SD = 0.58) (p < .0001) and men who reported no specific preference for gender expression (M = 3.21, SD = 0.65) (p < .001). Men with a masculine gender expression also differed significantly from men who reported no gender expression preference.

Mental Distress

Depression and anxiety were measured with two subscales of the Depression Anxiety Stress Scales (DASS; Lovibond & Lovibond, 1995). Each subscale consisted of seven items. Men were asked to report how frequently they had specific feelings during the past week (e.g., for depression: “I felt that life was meaningless”; and for anxiety: “I was worried about situations in which I might panic and make a fool of myself”); response options ranged from 1 = not at all to 4 = very much or most of the time. Cronbach’s alpha was .88 for depression and .83 for anxiety.

Sexual Minority Stressors

The following sexual minority stressors were included: discrimination while growing up, current discrimination, sexual identity confusion, and internalized homophobia.

Discrimination while growing up was measured using four items that asked: “As you were growing up, how often were you (1) made fun of or called names for being effeminate?; (2) hit or beaten up for being effeminate?; (3) made fun of or called names for being attracted to other men?; and (4) hit or beaten up for being attracted to other men?” (response options ranged from 1 = never to 4 = many times; Cronbach’s alpha: .79; adapted from Diaz, Ayala, Bein, Henne, & Marin, 2001).

Discrimination in the past year was measured using a previously validated scale adapted for this study (Herek & Berrill, 1992). Participants were asked to indicate the number of times in the past year that they were verbally insulted, physically threatened, had property damaged, objects thrown at them, been chased, spat upon, punched, hit, kicked or beaten, assaulted and sexually harassed because someone thought they were homosexual. We calculated the number of different kinds of discrimination men had experienced in the past year.

Sexual identity confusion was measured using four items; for example, “I’m not totally sure what my sexual orientation is” (response options 1 = disagree strongly to 6 = agree strongly; Cronbach’s alpha = .89; adapted from Mohr & Fassinger, 2000).

Internalized homophobia was measured using a previously validated 10-item scale (adapted from Mohr & Fassinger, 2006). Sample items include: “Sometimes I dislike myself for being a man who has sex with other men” and “I wish I were only sexually attracted to women” (response options 1 = disagree strongly to 6 = agree strongly; Cronbach’s alpha = .72).

Resilience Factors

As potential resilience factors, we assessed openness about one’s sexual orientation, gay community identification, and social support. Openness was measured using two items that asked how many of the men’s current heterosexual friends and casual acquaintances knew that they were sexually attracted to men (response options ranged from 1 = none of them to 5 = all of them; Cronbach’s alpha = .91).

We used a scale developed by Vanable, McKirnan, and Stokes (1998) to assess gay community identification. The scale includes four statements (e.g., “It is very important to me that at least some of my friends are bisexual or gay”); participants were asked to indicate the extent to which they agreed or disagreed with each statement (answer categories: 1 = disagree strongly to 5 = agree strongly; Cronbach’s alpha = .79).

Social support was measured using five items that asked how true it was that there was someone that men could rely on for money, food, a place to stay, to talk to if he has problems, to accompany him to the doctor, or help him if he gets hurt (response options ranged from 1 = always to 5 = never; total mean scores were reversed; Cronbach’s alpha = .86; Dandona et al., 2005).

Demographic Information and Sexual Orientation

The questionnaire also elicited demographics, including age, residential status, relationship status and duration, employment status, income, and religiosity. Three items were used to assess sexual orientation: (1) “Do you feel more sexually attracted to men or to women?”; (2) “In your current sexual fantasies, are you more aroused by men or by women?”; and (3) “Are your recent sexual experiences more with men or with women?” Response options were: 1 = only to women, 2 = mostly to women, 3 = to women and men equally, 4 = mostly to men, 5 = only to men. Cronbach’s alpha in the present study was .88.

Data Analyses

Multivariate analysis of covariance (MANCOVA) was used to assess differences between paired gender expression groups in mental distress, discrimination and other sexual minority stressors, and resilience factors. A MANCOVA with all dependent variables entered in the analysis was carried out for three comparisons: (1) the feminine versus masculine group, (2) the feminine versus neutral group, and (3) the masculine versus neutral group. This allowed us to use sets of controlling variables in each MANCOVA specific to each comparison, because the groups differed from each other on specific background variables, dependent upon which groups were compared. In the comparison between feminine versus masculine participants, we entered age, living in a township, being in an ongoing relationship with someone of the same-sex, being employed, and income as the control variables in the MANCOVA. We entered living in a township, being employed, and income as controlling variables in the MANCOVA comparing the feminine and the neutral group. Age was used as a controlling variable in the MANCOVA comparing the masculine and neutral group on the dependent variables. Wilks’ criterion was applied to determine statistical significance. When Wilks’ criterion was significant, we conducted analyses of covariance (ANCOVAs) to identify the variables on which the gender groups differed significantly.

Based on the established differences between the gender groups, we tested the assumption that the lack of differences in depression and anxiety between the three sets of subgroups was a consequence of the fact that the effect of supportive factors was canceled out by impeding factors. To test this, we conducted linear regression analyses for three subgroups: (1) the feminine and masculine men, (2) the feminine and gender-neutral men, and (3) the masculine and gender-neutral men, with depression and anxiety as outcomes and the stressors and resilience factors on which the respective subgroups differed significantly, as predictors. In these analyses, we used the same control variables as in the respective MANCOVAs.

Furthermore, we hypothesized that processes associated with mental distress might differ within the three gender expression groups. Using Model 1 for dichotomous moderators in the PROCESS program as developed by Hayes (2013), we tested this separately for each stressor and resilience factor that was significantly associated with anxiety or depression within each paired gender comparison (p < .05). Group membership (feminine–masculine, feminine-neutral, or masculine-neutral) was included as moderator. We controlled for those variables that were independently associated with the mental distress outcome within the paired gender groups. A significant interaction in the moderation analyses indicates that the association between a moderation variable (sexual minority stressors and resilience factors) and the dependent variable (distress or anxiety) is different for one or the other gender expression group. To interpret significant interactions, we evaluated simple slopes using methods described by Aiken, West, and Reno (1991). PASW Statistics 24.0 software was used to conduct all statistical analyses.

Results

Gender Expression, Sexual Minority Stressors, Resilience Factors, and Mental Distress

Feminine Men versus Masculine Men

A MANCOVA comparing feminine and masculine men on depression and anxiety, the four sexual minority stressors, and the three resilience factors as dependent variables showed a significant Wilks’ criterion (Wilks’ lambda = .80), F(9, 99) = 2.71, p = .007. In subsequent comparisons, we controlled for age, living in a township, being in an ongoing relationship with someone of the same-sex, being employed, and income. Feminine and masculine men did not differ significantly in depression and anxiety. With regards to sexual minority stressors, feminine men scored higher than masculine men on discrimination while growing up and lower on internalized homophobia. There were no significant differences regarding the other sexual minority stressors. In terms of resilience factors, feminine participants were more open than masculine men. Feminine and masculine men did not differ from each other regarding social support and gay community involvement.

Feminine Men versus Men Without Gender Preference

The MANCOVA comparing feminine participants with participants without preference showed a significant Wilks’ lambda (0.83), F(9, 93) = 2.20, p = .029 (Table 2). Subsequent ANCOVAs—controlling for living in a township, being in an ongoing relationship with someone of the same-sex, being employed, and income—showed that feminine participants and participants without a preference did not differ significantly in depression and anxiety. Regarding sexual minority stressors, feminine participants reported significantly higher scores on sexual identity confusion relative to participants without a preference. Other differences in sexual minority stressors between these two groups were not significant. Feminine men did not differ from the men without preference in terms of openness and gay community identification; they did, however, report lower social support than men without preference.

Table 2 Means and SDs for mental distress, sexual minority stressors, and resilience factors by gender expression and comparisons between groups (ANCOVAs)
Full size table

Masculine Men versus Men Without Gender Preference

The comparison between masculine and men without preference showed a significant Wilks’ lambda (0.77), F(9, 115) = 3.92, p < .001. Controlling for age, masculine men and men with no preference did not differ significantly in anxiety and depression. Masculine men scored higher than men with no preference on sexual identity confusion and internalized homophobia. Differences in discrimination while growing up and discrimination in the past year between these two groups were not significant. In terms of resilience factors, masculine men reported to be less open about their same-sex sexual attraction and to experience less social support, relative to men with no preference. Masculine men and men with no preference did not differ in terms of gay community identification.

Stressors, Resilience, and Mental Distress

Inspection of the observed differences between the three groups suggested to us that the lack of differences in depression and anxiety might result from the specific combination of sexuality minority stressors and resilience factors. For instance, the lack of differences in mental distress between the feminine and the masculine group could be the consequence of the fact that even though the feminine group had experienced more discrimination while growing up than the masculine group, they scored lower on internalized homophobia, which should bolster their mental health, while being more open about their same-sex sexuality, which as a resilience factor should contribute to mental health. We expected that when controlling for the factors on which each set of group differed, the group factor (e.g., feminine versus masculine) would show an independent association with the mental health outcomes. The outcomes of the linear regression analyses, presented in Tables 3 and 4, show, however, that this is not the case. None of the combined gender expression groups was independently associated with depression or anxiety. As far as sexual minority stressors and resilience factors were independently associated with mental distress in these sets of comparisons, the associations were in the expected directions. For instance, in the group combining feminine and masculine men, depression was independently associated with discrimination while growing up and internalized homophobia. Men who had experienced more discrimination while growing up reported more internalized homophobia and scored higher on depression. These findings made us wonder whether the sexual minority stressors and resilience factors had a differential impact on mental distress for each of the three gender expression groups.

Table 3 Association of sexual minority stressors and resilience with depression by gender expression comparisons (linear regression)
Full size table
Table 4 Association of sexual minority stressors and resilience with anxiety by gender expression comparisons (linear regression)
Full size table

Stressors and Resilience Predicting Mental Distress and Differences Within Gender Expression Groups

The moderation analyses showed that one of the nine interactions of the gender groups with the stressors and resiliency factors was significant, only for the feminine and masculine men. The association between depression and internalized homophobia was significantly different for feminine compared to masculine participants (p = .031). In feminine participants, internalized homophobia was positively associated with depression; however, this association was not significant for masculine participants (Fig. 1).

Fig. 1
figure 1

Unstandardized regression line for the association between internalized homophobia and depression, separate for Black South African MSM with a feminine and masculine gender expression

Full size image

Discussion

Assessing the association of gender expression as a categorical variable instead of a continuous variable, as we did in Sandfort et al. (2016), with mental distress produced findings in line with what was reported by Cook et al. (2013): (1) Men with a preference for a feminine gender presentation reported more discrimination while growing up compared to men with a preference for a masculine gender presentation, (2) among men with a preference for a feminine or masculine gender presentation, discrimination while growing up was associated with depression. However, (3) feminine men did not report more depression than masculine men. In fact, none of the comparisons between gender groups showed a significant difference in depression or anxiety. These findings support our earlier conclusion that the association between gender nonconformity and mental distress among gay and bisexual men is indeed more complex than generally assumed. Further analyses only partially helped us to understand this complexity. For instance, we did not find support for the assumption that higher scores on one stressor were canceled out by lower scores on another stressor or higher scores on a resilience factor when comparing the feminine with the masculine gender expression groups. We did, however, find that internalized homophobia had a differential impact on depression of feminine men compared to masculine men. We did not find a moderating effect of gender expression group for any of the other sexual minority stressors and resilience factors. It is not clear why internalized homophobia was only associated with mental distress in feminine men and not in masculine men. Given the number of comparisons, this could also be a chance finding.

The pattern of differences between the three gender expression groups in the sexual minority stressors and the resilience factors suggest, though, that a categorical operationalization of gender nonconformity is meaningful and that a preference for gender presentation is a critical factor in understanding differences in MSM populations (cf. Parker, Aggleton, & Perez-Brumer, 2016).

The elevated level of discrimination while growing up among feminine men compared to the masculine men is likely to be predominantly caused by their gender nonconformity (Skidmore et al., 2006). The fact that especially feminine men had experienced discrimination while growing up suggests that gender nonconformity is already present at a younger age and that, compared to masculine men, these men had a qualitatively different trajectory and process of coming out, in which gender nonconformity played a more crucial role.

Masculine men were less open and reported more internalized homophobia compared to both other groups. These men might be less open because, as masculine men, it is easier for them to pass as straight as their gender conformity does not raise any suspicion about their sexual orientation. In a social climate where rejection of same-sex sexuality is high, passing might be seen a safer option than coming out. It is possible, though, that not coming out precludes men from processing negative feelings about being attracted to members of the same sex and results in stronger internalized homophobia. Alternatively, elevated level of internalized homophobia could inhibit the masculine men from coming out (Weber-Gilmore, Rose, & Rubinstein, 2011).

Men without a preferred gender expression scored the lowest on sexual identity confusion and reported the strongest social support compared to the other two groups. It is likely that, compared to the feminine men, men with no preference have greater access to social support because it is not thwarted by the barrier of gender nonconformity. In comparison with men without preference, masculine men, being less open about their homosexuality, might have less access to social support that might support them in processing feelings of confusion about their sexual identity.

Demographic differences between the three groups help to further contextualize these findings. Feminine men were most likely to live in townships; it could be that the under-resourced township environment is less rejecting of femininity in men than the urban environment and that urban environment has supported defeminization among the two other groups of men. This would be congruent with Harry’s (1985) finding that persistence of gender nonconformity was stronger in gay men of lower social classes. Feminine men were the youngest, on average, and it is also possible that defeminization increases with age (Harry, 1983; Whitam, 1977). The finding that feminine men were most likely to be unemployed and had no or low income could be related to their age; it is also possible, though, that their gender nonconformity prevents them from finding and keeping jobs (Mahalik, Talmadge, Locke, & Scott, 2005).

The inspection of the pattern of differences between the three gender expression groups in the sexual minority stressors and resilience factors contributes to the further understanding of gender nonconformity in Black South African MSM. However, we were not able to solve the enigma identified in this population by Cook et al. (2013) that although gender nonconformity was associated with discrimination, that discrimination was associated with mental distress, but that gender nonconformity and mental distress were not associated.

As discussed before (Sandfort et al., 2016), limitations of this study include its cross-sectional design and the use of self-report. In addition, even though our gender expression assessment has construct validity, it is not clear what men’s expression preference implies in social life. Of critical importance is whether their expression preference is perceived as such by others. The higher level of discrimination suggests, however, that this is indeed the case. In addition, it is not clear what the content validity is of the scale that assessed current discrimination; although more sophisticated scales are available, it is not clear what forms of discrimination are most prominent in the South African setting. Given the associations we reported for discrimination with depression and anxiety for the total sample (Sandfort et al., 2016), we assume that our measure picked up some of the relevant forms of discrimination. Furthermore, because our design only included MSM, we are not able to establish whether mental distress in this population is elevated in comparison with men who exclusively have sex with women.

An additional limitation needs further discussion: we did not ask participants how they identified in terms of their sexuality and gender. That question was not included because we were of the opinion that responses to such a question are meaningless, if it is not clear what labels such as gay, bisexual, and transgender mean to study participants (cf. Sandfort & Dodge, 2009). We thought that assessing sexual orientation in terms of attraction would be more informative.

We regret this decision: inclusion of identity questions could have ascertained whether the three gender expression categories reflect a critical distinction among the participants in that feminine MSM would be more likely to identify as transgender (and thus should have been labeled trans women in this study), masculine MSM would be more likely to identify as bisexual, and men without would be more likely to identify as gay. If this were the case, it could help to better understand the relationship between gender nonconformity, discrimination, and mental health.

However, another study in the same population made clear that the adoption of the label transgender is rare (Sandfort, Lane, Dolezal, & Reddy, 2015); with only a few exceptions, all gender-nonconforming participants in this study identified as “gay.” It could be that even though from a Western perspective at least some of these gender-nonconforming men would be considered “transgender,” these men do not do so because of unfamiliarity with the label and its meaning. The history of the transgender movement in the U.S. includes a moment in which transgender persons perceived themselves as “gay,” the gay community probably offering the first opportunity for identification (Minter, 2000). If this reasoning applies to Black MSM in South Africa, it is likely that when the concept of transgender diffuses, a specific transgender category will emerge, separate from, although quite likely still associated with the gay community. This reasoning is supported by the establishment in 2010 of a transgender organization in the province that included our study site, named TIA (Transgender and Intersex Africa) as well as the description of experiences of transgender persons in the same province (Husakouskaya, 2017). It is also possible that these gay men prefer a feminine gender presentation based on their sexual practices, equating a stronger interest in receptive anal intercourse with being feminine.

Subsequent research with this population should not only include sexual and gender identity questions, it should also explore familiarity with these terms. In addition, such studies should assess gender nonconformity during childhood. If men with varying gender expressions already differ while growing up, it is likely that their sexual identity development process varies, eliciting differing social responses and enabling different coping strategies. Including childhood gender nonconformity in research would allow us to further disentangle the complex relationship between gender nonconformity and mental health in this population.