Weinrich, James D., Atkinson, J. Hampton, Jr., McCutchan, J. Allen, and Grant, Igor (1995) Is gender dysphoria dysphoric? Elevated depression and anxiety in gender dysphoric and nondysphoric homosexual and bisexual men in an HIV sample. Archives of Sexual Behavior 24: 55-72. Copyright 1995 Plenum Publishing Corporation

Is Gender Dysphoria? Elevated Depression and Anxiety in Gender Dysphoric and Nondysphoric Homosexual and Bisexual Men in an HIV Sample


Although most cases of HIV infection in North America have been acquired sexually, only a few sex researchers have applied their unique perspectives and skills to the AIDS epidemic, and few mainstream psychiatrists and psychologists have addressed the sexual aspects of the disease. That is, there has been little cross-fertilization of ideas between the psychiatric research community and the sex research community.

The present report is an attempt to introduce what might be called a traditional sexological perspective into a mainstream psychiatric research program on HIV disease and neuropsychology, with results that may be pertinent to questions of interest to a broad range of scientists and clinicians. The HIV Neurobehavioral Research Center (HNRC) is a large, federally funded interdisciplinary research project affiliated with the Department of Psychiatry at the University of California, San Diego. Its basic mission is to investigate the natural history and prevalence of the neuropsychological sequelae of HIV infection, with attention given to neurology, psychiatry, virology, medicine, brain imaging, and other disciplines.

A central concern of the HNRC is the assessment of rates of psychiatric morbidity such as depression and anxiety. We have found that the levels of depression and anxiety in our homosexual subjects, whether HIV positive or HIV negative, are substantially higher than those found in representative general population samples. For example, our homosexual populations (HIV-positive and HIV-negative) had lifetime frequencies of major depression (diagnosed by the Diagnostic Interview Schedule for DSM- III of 31-32% (N = 135), compared to 9% in 22 HIV-negative heterosexual controls (Atkinson et al., 1989). Similar findings have been presented from other investigators of AIDS-related homosexual samples (Hays et al., 1990; Perry et al., 1990; Viney et al., 1991; Williams et at., 1991), although the finding is not universal.

Of course, AIDS is a life-threatening disease arguably capable of triggering depression in and of itself. It could do so indirectly through the conscious recognition of one's increased likelihood of mortality due to the disease, or directly via hypothesized effects on neurotransmitters or other forms of CNS morbidity. However, some of the studies that found high levels of depression in HIV-related samples of homosexual men reported that the history of depression predated the AIDS epidemic (Atkinson et al., 1988; Joseph et al., 1990; Williams et at., 1991). Accordingly, the higher levels of depression now observed can probably not be explained merely as the result of the undoubted pressures caused by this epidemic.

Although it might in theory be expected that homosexuals living in a hostile society may experience depression as a reaction to the social stigma attached to their orientation, in fact there is little if any documentation of such an increased level of rigorously diagnosed depression from nonclinical samples studied before the AIDS epidemic (Saghir and Robins, 1973; Gonsiorek, 1991). We found two reports of higher depression from this period, but these studies did not assess depression with a clinical interview or standardized questionnaire. One found higher depression among homosexual men (compared with the general population) in samples from three different countries (Weinberg and Williams, 1974, p. 152), and higher levels among effeminate homosexual men than among those who were not effeminate (p. 172). Although this result corresponds well to those we report presently, that study's measure of depression was the subjects' responses to an item asking if they "felt they were happy" persons. The other study used subjects' answers to seven questions likewise not previously standardized (e.g., how frequently did they feel life was worthwhile, experience hurt feelings, etc.: Bell and Weinberg, 1978, pp. 200-201). Even the latter study found that levels of depression were not elevated among the male homosexuals who were in monogamous relationships or who were "functional" (i.e., uncoupled but functioning successfully in the gay world). One notable study from this period which used more rigorous psychiatric methodology found no difference in levels of depression by sexual orientation (Saghir and Robins, 1973, pp. 113- 118).

Accordingly, it is important to ascertain which aspects of sexual orientation are associated with depression in homosexual populations, if any. Given the unfortunate history of the classification of homosexuality per se as a disease, it is worthwhile to study depression in this population without presupposing that it is a reflection of an underlying pathology but might instead be a response to environmental circumstances or even an intermittent aspect of otherwise normal personality in certain personality types.

One of the few psychological traits that has been shown, in both retrospective and prospective studies, to be both statistically and substantively significant when differentiating and contrasting the average backgrounds of heterosexual and homosexual men has been termed childhood gender nonconformity (CGN) by some authors (Bell and Weinberg, 1978), childhood femininity by others (Green, 1974, 1987; Freund et al., 1974), and gender dysphoria in the DSM-III-R. Such behaviors are far from universal in the histories of homosexual men in Western societies, but they are not uncommon, and they are apparently much less common among men who identify themselves as heterosexual. One questionnaire used to assess such behavior is the Freund Feminine Gender Identity questionnaire (FGI; Freund et al., 1974, 1977).

We have observed a mild correlation between homosexual men's scores on the FGI and their interest in receptive anal intercourse (Weinrich et al., 1992). (Note that we did not assume a priori that sex-role nonconformity such as that measured by the FGI must necessarily be related to variability in sexual orientation or preferences for particular erotosexual roles; we observed that a moderate correlation existed.) This correlation alone should alert AIDS researchers that CGN may be a personality variable that contributed to HIV exposure (before the era of safer sex), which might in turn suggest that HIV-related samples are likely to have a higher proportion of high-CGN men than more widely representative samples. And it should alert researchers that CGN could be important to study whenever sexual orientation is investigated.

Logically speaking, a desire to be a member of the opposite sex (whether in adulthood or childhood) is not necessarily dysphoric. Anyone expressing a desire to change an important facet of their life might be suspected of being dysphoric about their current circumstances, but whether this presumed dysphoria actually exists, can be detected by our measurement instruments, and ever reaches the level of clinical depression are empirical questions. Accordingly, we thought it important to assess the extent to which CGN, as measured by FGI, is in fact associated with dysphoria and to that extent might account for some of the differences in depression observed between heterosexual and homosexual male samples. Note that one of the early papers on the FGI found that homosexual psychiatric patients scored higher on the FGI than did a group of homosexual nonpatients (Freund et al., 1977, p. 512).

If it is indeed the case that the previously cited samples of homosexual men from before the AIDS epidemic show no evidence of clinically depressed mood, but most samples of HIV-positives and HIV- negatives report high levels of a lifetime diagnosis of major depression perhaps dating from before the epidemic, then we face some troubling and intriguing questions. Are sample selection biases before and after the advent of AIDS so different that they could account for such a striking difference in findings about depression? This is counterintuitive, albeit possible. Are homosexual men who are currently depressed by the HIV health crisis likely to retroactively distort their accounts of lifetime depression? This would challenge well-validated clinical interviewing instruments. Societal pressures and stigmatization attached to homosexuality might account for a proportion of the elevated lifetime rates of major depression, but there seems to be little evidence that this stigmatization is higher now than it was when the previous studies were conducted (indeed, it seems likely that just the opposite would be the case, given the progress made in struggles for gay rights). However, individual developmental factors, which are now just becoming measurable, may add to our understanding of this process.

The purpose of the present paper is to see whether anxiety, depression, and their related symptoms might be illuminated by measuring CGN within a homosexual and bisexual sample. Our specific hypothesis is that men with high FGI scores would show elevated average levels of anxiety and depression. If confirmed, this hypothesis would predict that samples of homosexual men varying in average CGN could easily have different average levels of syndromic depression and/or depressed or anxious symptoms (all of which would have predated the AIDS epidemic). If there were reason to expect that AIDS-related samples of homosexual men would be skewed toward individuals of higher CGN, then this hypothesis would receive additional support.



The sample was drawn from a group of 441 male subjects from the HNRC's total subject pool. Of these men 55 had described themselves as lifelong heterosexuals; of the others, 127 had not completed the FGI (or their questionnaires were unusable) and 3 more had FGI scores but none of the psychiatric variables available. These men were excluded from the present sample. Also excluded were one transsexual individual (with an FGI score of 46, 10 points higher than any other subject in the study at the time) and the subject with the next highest FGI score - who had been excluded as an outlier in the analyses that produced the factor scores used in the present analyses (Weinrich et al., 1992). [TABULAR DATA OMITTED]

Table I lists the demographic characteristics of the resultant 254 men in the sample, as well as indicating the sample sizes for the various procedures. Typical participants had some college education, were in their 30s at time of entry into the study, were Caucasian non-Hispanics, and had FGI scores with roughly the same distribution as those from the original validation sample (see Freund et al., 1977, Fig. 3). However, it was impossible to assess this similarity quantitatively, because that paper did not report actual scores -only a graphic line chart of the distributions, with each pair of adjacent scores grouped together for smoothness in charting.

The HNRC sample was recruited from the membership of a previous longitudinal study of HIV in San Diego, by advertisements placed in local newspapers, by word of mouth, and by referral from the HIV care facilities associated with large local hospitals. Inclusion criteria for the HNRC were: male, ages 18-49 at entry, and having more than 9 years of education. Exclusion criteria were: history of alcohol or other psychoactive substance use disorders in the past 12 months, intravenous drug use (ever), clinical diagnosis of AIDS dementia complex (ADC), chronic medical illness (e.g., diabetes mellitus, chronic obstructive pulmonary disease), or neurologic disorder unrelated to HIV (e.g., head trauma).


Subjects were screened in face-to-face interviews and signed a consent form. Over the course of the larger study, they completed two questionnaires: the FGI scale (Freund et al., 1977) and a self-report measure of psychiatric symptoms, Profile of Mood States (POMS, McNair et al., 1980). They also underwent a standard, face-to-face diagnostic psychiatric interview, the Structured Clinical Interview for DSM-III-R (SCID, Spitzer et al., 1988), and in some cases had already completed another, similar interview, the Diagnostic Interview Schedule for DSM-III (DIS, Robins et al., 1980). Several other self-report measures of depressive and anxious symptoms were also administered to the sample, but for reasons of space we have chosen only the instruments for which the largest N is available. (Results are essentially the same for all measures.) We included the DIS in the present report because it is the only instrument we used that provides a diagnosis of Ego-Dystonic Homosexuality (dropped in the revision of DSM-III to DSM-III-R).

Data were entered into computers, scored, and analyzed using JMP version 2.0 from the SAS Institute. Stepwise multiple regressions were performed by the StatView program. The usual subscores from the POMS were calculated; we discuss only Depression/Dejection and Tension/Anxiety in the present report. The two POMS variables have continuous distributions, and they were first correlated with the FGI total score in a series of simple regressions. The two variables with categorical distributions - DIS Ego-Dystonic Homosexuality and SCID Lifetime Depression - were first correlated with the FGI total score in simple logistic regressions.

When results (see below) confirmed the hypothesis that these variables were related to FGI, exploratory analyses were undertaken in which the dependent variables were correlated with 7 aspects of the FGI: FGI total score, FGI A, FGI B, and four FGI factor scores. FGI Part A consists of the total score on the first 19 items of the FGI (mostly childhood items), whereas Part B consists of the total score on the latter 10 items (mostly adult gender-identity items; note that what we term Part B was split in the original description of the FGI into Parts B and C). The four factor scores had been determined by a previous factor analysis (Weinrich et al., 1992), and cover the following areas. The Masculine Role factor is most heavily influenced (loaded upon) by FGI items concerning childhood interest in stereotypical adult masculine occupations and hobbies (although, like all FGI subscales, it is scored so that high values reflect high levels of feminity). The independent Feminine Role factor is most heavily loaded upon by FGI items concerning such interest in adult feminine sex-typed occupations, as well as by items concerning childhood play preferences for girls' toys or girls as playmates. The Core Gender Identity factor is most heavily loaded upon by items stating that one wished to have been born a girl, wants to be a woman now, and so on. The Sexual Desire factor is most heavily loaded upon by items relating to sexual desires stereotyped as "feminine" in our culture, such as preference for masculinity in a sexual partner and preference for an insertee role in sexual behavior. Details of the factor analysis have been published in this journal (Weinrich et al., 1992).

Of course, some of the independent (FGI) variables are not independent of each other (although the four factor scores were, by design, independent in the sample used in the factor analysis); nor are the dependent (psychiatric) variables independent of one another. Accordingly, the entries in the tables giving results should be interpreted with caution. Nevertheless, the patterns in those tables tell a consistent story, which we examine in the Results.

After the simple correlations between FGI and the psychiatric variables were performed, we undertook a series of regressions in which the various psychiatric variables were correlated with the subjects' HIV sero-status. We also performed multiple regressions in which the psychiatric variables were regressed against the independent variables of FGI total score and CDC stage (rated as zero for HIV-negatives). These regressions addressed whether stage of disease might influence our depression and anxiety measures. Next, we conducted a series of regressions to assess whether age was a significant correlate of the measures of depression and anxiety in our sample. Then we used stepwise multiple regression to help us decide which of the seven FGI measures were most important in explaining our correlations. Since these seven variables are not all independent of each other (FGI total is the sum of FGI A and FGI B, for example), we conducted a second stepwise regression using only the four factor scores (which were automatically designed to be independent of each other in the study which identified them, Weinrich et al., 1992), in the hopes that this would identify which "types" of femininity were associated with depression and anxiety in men.


Table II shows the correlations of the FGI and its subscores with the two POMS subscales, whereas Table III shows results of corresponding logistic regressions for the SCID and DIS diagnoses. It can be seen from the first row in these two tables that the total FGI score had significant (p < 0.01 or 0.001) associations with each psychiatric variable except Ego-Dystonic homosexuality (p > 0.25). Accordingly, our hypothesis that high FGI scores would be associated with higher levels of depression, depressive symptoms, and anxious symptoms was confirmed.

In further analyses, some intriguing patterns emerged from the breakdowns of FGI into Parts A and B, and into the four factor subscores. FGI Part A correlated significantly with the same three variables with which the total score correlated (i.e., every one except Ego-Dystonic Homosexuality). Part B correlated, at weaker strength, with the depression measures but not significantly with POMS Tension/Anxiety. We also have data on the recency of the DIS diagnosis of Major Depression (which we do not report in the tables). Eight of the 9 individuals who were diagnosed as depressed (by the DIS) in the 2 weeks prior to their baseline visit scored at or above the median for FGI B; this fact explains over 20% of the variance in a logistic regression of current Major Depression by FGI Part B! Although this is a finding concerning a small number of depressed subjects, it is statistically highly significant (p < 0.0005) and deserves further exploration. 

    Table II. Pairwise Regressions of FGI Components with POMS Stores(a)

MS scores
    FGI components           Depression/Dejection     Tension/Anxiety
    Total                          .247(g)            &
nbsp;   .212(g)
    Part A                         .232(g)             
    Part B                         .156(e)             
    Masculine Role                   -(b)                  -(b)
    Feminine Role                  .111(c)                .126(e)
    Core Gender Identity           .204(f)                .150(d)
    Sexual Desire                    -(b)                  -

    a Numbers in the body of the table are the square root of the
    proportion of variance explained by the regression, namely, the
    regression coefficient r.
    b p < 0.10.
    c p < 0.10.
    d p < 0.05.
    e p < 0.02.
    f p < 0.01.
    g p < 0.001.

    Table III. Logic Regressions of DIS and SCID by FGI(a)

                        SCID Major Depression     DIS Ego-Dystonic
    FGI components            (Lifetime)            Homosexuality
    Total                      .169(f)                &
nbsp; -(b)
    Part A                     .156(f)                 
    Part B                     .118(d)                 
    Masculine Role             .113(c)                  -(b)
    Feminine Role                -(b)                   -(b)
    Core Gender Identity       .136(d)                .194(e)
    Sexual Desire                -(b)                   -(b)

    a Numbers in the body of the table are the square root of the
    proportion of variance explained by the regression. In a logistic
    regression, there is no real correlation coefficient r, but the
    square root of R squared is provided for a comparison with Table II.
    The direction of the correlation in each of the significant
    associations is similar: High FGI scores tend toward higher
    depression and more frequent ego-dystonic homosexuality.
    b p < 0.10.
    c p < 0.10.
    d p < 0.05.
    e p < 0.02.
    f p < 0.01.

Among the four FGI factor scores, the Sexual Desire factor never achieved even the 0.10 level of significance, and the Masculine and Feminine Role factors rarely did (only once barely bettering a p of 0.05). It was the Core Gender Identity factor that was apparently responsible for most of the correlations (for further discussion of this point, see below); incidentally, it was the only FGI component which correlated significantly with Ego-Dystonic Homosexuality. This probably says more about the explanatory power of this FGI factor than it does about the psychiatric diagnosis. 
    Table IV. Seropositivity Status by Psychiatric Variables

                                          n &
nbsp;       R.sq.         p
    Continuous dependent variables (by t test)
    POMS Depression/Dejection            244       .003        0.3794
    POMS Tension/Anxiety                 244       .008        0.1523

    Dichotomous dependent variables (by Pearson Chi sq.)
    SCID Major Depression (lifetime)     202       .005        0.4600
    DIS Ego-Dystonic Homosexuality       120       .002        0.5250

Table IV shows that none of our psychiatric variables correlated significantly with seropositivity status; all p values are > 0.15 and the variance explained is less than 1% in each case.

Results of the multiple correlation analyses of the psychiatric variables against CDC stage and total FGI score are presented in Table V. For the POMS subscores, the conclusions are unambiguous: The model accounted for a statistically significant proportion of the total variance, the lack of fit probabilities were suitably high (confirming that there was no pattern suggesting that further variables needed to be entered into the model), the FGI coefficient was significantly different from zero, and there was no statistically significant effect of CDC stage.

For SCID Major Depression, the multiple regression results (bottom half of Table V) are less clear, because the model p as a whole was only a trend (0.05 < p <= 0.08). However, it was reassuring to note that the p value of the FGI coefficient was significantly different from zero, and that the coefficient for the CDC stage was not. For Ego-Dystonic Homosexuality, there were no significant associations found, echoing the results of the univariate analyses. The result reported in Tables IV and V provide no evidence that the univariate associations between the FGI and the psychiatric variables are mere artifacts of HIV status. [TABULAR DATA OMITTED]

Results of the regressions of the psychiatric variables by age are presented in Table VI. For the sample as a whole, the only significant correlation is with Ego-Dystonic Homosexuality, and it is in the negative direction: Older men report less Ego-Dystonic Homosexuality than younger men. A breakdown of the sample into HIV-positive and HIV-negative subjects shows no significant relationships in the larger, HIV positive subsample. On the other hand, younger HIV-negative men are more likely to have higher POMS Depression/Dejection (p < 0.03) and higher rates of Ego- Dystonic Homosexuality (p < 0.0001).

Results of the stepwise multiple regressions are presented in Table VII. When all 7 FGI variables (FGI total score, FGI A, FGI B, and the four factor scores) were available to the multiple regression, only one variable significantly entered the regression in each case: twice it was the FGI total score, and once each it was FGI A and the factor FGI Core Gender Identity. In the second analysis using only the four factor scores, the results were striking: FGI Core Gender Identity entered each of the four regressions first, and was the only significant variable in three of the four cases.



In our sample of homosexual and bisexual men, those with highly gender nonconforming FGI scores were more likely to exhibit current depressive or anxious symptoms, and to have received lifetime diagnoses of major depression. Childhood gender nonconformity (FGI Part A) was usually a better predictor than adult gender nonconformity (Part B). A notable exception came in predicting current DIS-diagnosed depression, in which FGI Part B explained over 20% of the variance, in the admittedly small group of subjects who were depressed at baseline interview. When the FGI was broken into four subscales using a previous factor analysis, it was the Core Gender Identity factor that seemed responsible for most if not all of the substantial correlations. We found no evidence that HIV seropositivity, stage of HIV disease, or age could account for these results, even though increasing age might be expected to increase the chance that one would report a depressive episode.

These results apparently confirm the early report on the FGI which demonstrated that a group of "homosexual patients scored significantly higher on feminine gender identity than did ... [a group of] homosexual nonpatients.... A breakdown showed that the patients differed on items related to cross-dressing, to the desire to be a woman, and to the desire for a strong masculine partner" (Freund et al., 1977, p. 512). The patients had been referred for "difficulties in social or sexual adjustment." Although no breakdown by diagnosis is provided, such a result is exactly what we would predict if much of the psychiatric morbidity in the patient sample consisted of depression or anxiety.

Just as important in an exploratory study such as this are the possible associations that were not significant. Ego-Dystonic Homosexuality, a DSM-III diagnosis dropped in DSM-III-R, correlated significantly with only one of the seven FGI components under study. Incidentally, Ego-Dystonic Homosexuality correlated (in analyses not reported in this paper) with only two of eight other measures of depression or depressive symptoms and none of anxiety. This suggests that such an ego-dystonic state is not very dysphoric, with the implication that this diagnosis is of doubtful validity.

On the other hand, we have fairly strong evidence that gender dysphoria is dysphoric: It was the Core Gender Identity factor of the FGI that best accounted for the elevated levels of anxiety and depression seen in our sample. Note that none of the items loading heavily on this factor are explicitly worded to elicit dysphoric feelings. That is, a typical question is worded, "Between the ages of 13 and 16, did you wish you had been born a girl instead of a boy often, occasionally, or never?" (as opposed to, let us say, "Between the ages of 13 and 16, were you unhappy that you had been born a girl instead of a boy?"). Although we would prefer, for theoretical and philosophical reasons, to avoid prejudging the nature of the cross-gender feelings by choosing a term like "dysphoria," we feel that the term is easier to justify in this case empirically than "dystonic" could be in "ego-dystonic homosexuality."

Since our subjects are homosexual or bisexual, we have presented no evidence directly relevant to the question of whether homosexual orientation per se is related to depression or anxiety. However, the FGI Sexual Desire factor, which is loaded upon by FGI questions assessing receptive (so-called "passive") sexual behavior, had no significant correlations with our depression or anxiety measures. This casts doubt, if any is needed, on the hypothesis that some homosexual mens' interest in such sexual behaviors leads inevitably to psychopathology--at least depressive or anxious psychopathology.

Since all four aspects of childhood femininity in boys are discouraged by typical Western socialization patterns, but only one of the four aspects correlates consistently with depression and anxiety, let us discuss some possible reasons why the Core Gender Identity factor may have emerged so strongly in these analyses. Consider a boy with a budding interest in submissive sexual behaviors, or with an interest in stereotypical feminine roles or occupations, or with a clear disinterest in so-called masculine roles or occupations. In each of these three cases, there is an interest that is disapproved of by society at large. Note that a boy with any combination of such interests can grow into a man who can realistically satisfy those interests--by leading a certain kind of homosexual life-style, by avoiding occupations such as auto mechanic, or by becoming a dancer or dress designer (these examples are taken from FGI items loading on the corresponding subscales). Our results suggest that such a boy can avoid the anxiety and depression that might otherwise be hypothesized to result from social disapproval of such interests and activities. But now consider a boy with a strong desire to have been born a girl. Such a boy cannot truly grow into an adult woman, and so these wishes cannot realistically be fulfilled; even a transsexual operation only satisfies a portion of the wish. Our data suggest that such boys often have not avoided anxiety and depression, and may be at a relatively high risk of recurrent depression as adults.

This insight has both theoretical and clinical significance. If therapy is offered to such gender-dysphoric boys, our data suggest that there may be some benefit to trying to help them become less dysphoric in terms of their core gender identity. On the other hand, there may not be a measurable benefit, in terms of adult psychological adjustment, in trying to change boys' interests in feminine roles and occupations, their disinterest in masculine roles and occupations, and their later interest in sexual relations with men.

Theoretically, this insight is important in the following ways. First, it suggests that there may be at least two types of routes to a homosexual orientation: One type of route begins with gender dysphoria, the other does not. Second, although our data do not bear directly on the question of nature or nurture in the cause of depression or homosexual orientation, they illuminate and constrain each point of view. For example, one explanation of the higher levels of depression seen in the gender-dysphoric portion of our sample is that such men become depressed because they want to become something they cannot truly become, whereas men who are gender nonconforming in other ways want to do things that men can, in fact, do (and succeeded at doing), even if much of society disapproves of men doing those things. A more psychohormonal explanation of our finding would start with the presumption that core gender identity is in part a product of genetic and/or hormonal factors operating typically in genetic females to produce a feminine gender identity, and atypically in genetic males to produce gender dysphoria, and that whatever it is that produces higher levels of depression in women as opposed to men will produce higher levels of depression in gender-dysphoric males than in nondysphoric ones. Note that this latter explanation takes no position on the mechanism that one assumes is responsible for the higher lifetime rates of major depression in women observed in many modern Western populations.

All available data support the conclusion that groups of homosexual men tend to have more individuals with high FGI and related femininity than groups of the heterosexual men; this is true not only in our own sample (results not presented here) but also in others (Bell et al., 1981; Green, 1987), including those with which the FGI was developed (Freund et al., 1977). The fact that high FGI scores are apparently related to a preference for anal intercourse as insertee, first empirically demonstrated only recently (Weinrich et al., 1992), suggests that many AIDS-related samples of homosexual men usually recruit subjects with higher average levels of FGI than other more representative samples - because receptive anal intercourse is the most common route of HIV infection for homosexual men. This in turn suggests that AIDS samples of homosexual men will have disproportionately high levels of depression, with the depression even predating the AIDS epidemic.

It could also explain why HIV-negative subjects would have high levels of depression, also predating the AIDS epidemic; one need merely hypothesize that our HIV-negative control sample has a healthy representation of high-FGI men who have avoided HIV exposure. (In analyses not reported here, we calculated logistic regression lines for SCID and DIS depression diagnoses as a function of FGI score, in each CDC group. These regression lines crossed the intercept of FGI = 0 at a frequency of the depression measure of 10% or less in the seronegative group, suggesting that low-FGI homosexual men have ordinary levels of depression.) It fails to explain, however, why earlier studies of depression in men who must have been HIV-negative did not have elevated levels of depression in comparison with heterosexual control groups. The obvious possibility--that self-selection biases have often differed before and after AIDS came upon the scene--is an attractive hypothesis which we cannot test with the present data (since we do not yet have appropriate control groups), but which remains somewhat puzzling. Accordingly, and unfortunately, we cannot claim that we have solved the puzzle we stated in the Introduction.

Yet our results suggests that it is important to control for--or at least describe--the amount of FGI (or some other measure of femininity) when comparing different populations, even if each of the populations is composed entirely of homosexual men. Anecdotal and clinical evidence suggests that different sources of subjects (e.g., different gay bars) could produce samples with different levels of average FGI. At least one study found that homosexual psychiatric patients had higher FGI scores than homosexual nonpatients, although there was no breakdown to indicate the diagnoses pertaining to those patients (Freund et al., 1977). Another study found that most of the adjustment differences by sexual orientation in a sample of homosexual and heterosexual men disappeared when the comparison was restricted to men scoring low on the Gough femininity scale, although that study found no masculinity-femininity difference on the Depression subscale of the Scheier and Cattell Neuroticism Scale Questionnaire (Siegelman, 1972).

We conclude that in homosexual and bisexual men, high levels of core gender-identity nonconformity are associated with higher levels of depression, and with higher levels of depressive and anxious symptoms, in comparison with gender conforming men. Our work adds to a growing body of evidence that homosexual men are not a homogeneous group, and additionally suggests that feminine homosexual men do not constitute a homogeneous group, either. Clinicians and researchers interested in AIDS or sexual orientation issues should be aware of this heterogeneity and its implications.


We thank Ray Blanchard for suggesting the theme, "Is gender dysphoria dysphoric?", Andres Sciolla and Sidney Zisook for comments, and an anonymous reviewer for the detailed reference to Weinberg and Williams (1974).


COPYRIGHT 1995 Plenum Publishing Corporation