Predictors of movement toward homosexuality: a longitudinal study of bisexual men.

Despite recent books (Firestein, 1996; Garber, 1995), articles (Fox, 1995; Leland, 1995), and increased exposure on television talk shows, bisexuality as an enduring sexual orientation is not generally well accepted. In a recent survey of mostly gay-identified male readers of The Advocate, about one third of the respondents did not "believe in bisexuality" (Lever, 1994). About 40% of the respondents to this survey said they had identified as bisexual en route to seeing themselves as gay. Several theorists (Cass, 1979; Coleman, 1987; Gagnon, 1977; Klein, 1978; Troiden, 1989) have speculated that at least some self-identified bisexual men will eventually see themselves as gay. In a society where male-to-male sexual behavior is stigmatized, adopting a bisexual identity may provoke less anxiety than adopting a homosexual identity. Seeing oneself as bisexual may allow one to retain a feeling of commonality with heterosexuals and may assuage the alienation and isolation felt by many men as they become aware of their sexual attraction to other men.

On the other hand, bisexuality is becoming more and more accepted as a "legitimate" and enduring sexual orientation (Coleman, 1987; Fox, 1995). Forty percent of the respondents in the Advocate survey agreed that some men really are attracted to both men and women. Probably some behaviorally bisexual men will in time move toward a homosexual self-identity, whereas others will have sexual feelings for both men and women, and/or they will engage in sexual behavior with both men and women for much or all of their lives. Longitudinal data from a small sample (N = 30) of self-identified bisexual men supported this speculation (Stokes, McKirnan, & Burzette, 1993). About 40% of the men they followed for one year moved toward the homosexual end of selfrated sexual feelings. Sixty percent either did not change over the one-year period or they moved toward a more heterosexual orientation. Although conclusions based on data from this small sample of convenience should be regarded with caution, the results suggested that some self-identified bisexual men will eventually see themselves as gay; for others, the bisexual self-identity will endure.

There is a paucity of empirical research on variables that distinguish men who remain stable in a bisexual orientation from those who move toward homosexuality. Stokes et al. (1993) reported some interesting trends within their sample of bisexual men. Respondents who moved over time toward more homosexual self ratings, compared to those who did not show such movement, reported at the initial interview more self-homophobia, anxiety, hostility, and interpersonal sensitivity. The authors hypothesized that these men were in transition to a more homosexual identity and were finding the process stressful. These data are consistent with theories of homosexual identity formation (Cass, 1979; Coleman, 1982; Dank, 1971; Friend, 1991; Minton & McDonald, 1984; Troiden, 1979, 1989), which have suggested that the early stages of the development of a homosexual identity are characterized by feelings of confusion, alienation, anxiety, and depression that result from an internalization of society's stigma surrounding homosexuality. Individuals who are moving toward a homosexual identity are likely to experience feelings of self-homophobia, anxiety, and depression as they try to resolve the incongruence between their sexual feelings and society's message about homosexuality. Negative affect seems likely for these men, at least compared to men who have no same-sex attraction or sexual experience. Among behaviorally bisexual men, however, those who eventually identify as homosexual may not experience more negative affect than those who remain stable in a bisexual orientation, because same-sex behavior is common to both.

In this study we examined psychosocial correlates of movement toward homosexuality. We acknowledge that the phrase movement toward homosexuality is imprecise; change in self-ratings' toward the homosexual end of a bipolar scale of current overall sexual feelings is a more precise reference for the construct we are examining. In addition, we realize that homosexuality is not some monolithic construct one moves toward or from in a linear way; movement toward homosexuality fails to capture the fluid and contextual nature of sexuality. We also acknowledge that changes in sexual feelings and orientation over time occur in all possible directions. Nonetheless, we use the phrase movement toward homosexuality for its economy of expression, and we focus in this study on that particular type of change over time.

The use of alcohol and other substances may help alleviate the anxiety that is associated with developing a homosexual identity (Coleman, 1982; Peterson, Ostrow, & McKirnan, 1991; Troiden, 1989). Substance use allows individuals to avoid focusing on feelings and behaviors they see as unacceptable. Hence, individuals in the process of transition may be more likely to use substances with their same-sex partners because it helps them to cope with the anxiety surrounding such behavior. On the other hand, one could argue that the men who are "just fooling around" when they engage in homosexual activity--that is, those who will not move toward a more homosexual self-identity--are more likely to use substances and that the substance use is causally related to their same-sex behavior.

Consistent with the psychological truism "the best predictor of behavior is past behavior," movement toward a homosexual identity and exclusively same-sex behavior may be predicted well by the number of homosexual experiences and/or a relative absence of heterosexual experiences. In a sample of 150 self-identified homosexual men, Troiden and Goode (1980) discovered that the fewer high school heterosexual experiences a respondent had, the younger he was when he first suspected he might be gay, labeled his feelings as gay, and labeled himself as gay. More high school homosexual experiences were also associated with earlier gay self-identification. Troiden (1989) proposed that adolescent homosexual experiences and an absence of heterosexual experiences expedited the process of homosexual identity formation because of the perceived (and probably actual) link between these characteristics and adult homosexuality.

Of course, some men with a significant amount of heterosexual experience eventually develop a homosexual identity. Furthermore, some men who eventually identify as gay may attempt to inhibit homosexual behavior in the early stages of their transition because such behavior arouses anxiety (Cass, 1979;Troiden, 1989). Controlling homosexual behavior is certainly easier than controlling homosexual desires or fantasies. Therefore, fantasies may be a better predictor of later movement toward a homosexual identity than is same-sex behavior. Despite these qualifications, in a group of behaviorally bisexual men, increasing numbers of homosexual experiences and a relative absence of heterosexual experiences should predict eventual self-identity as homosexual. Similarly, having a romantic relationship that includes emotional attachment as well as sexual activity with a man and not having such a relationship with a woman should be associated with eventual self-identity as homosexual. In addition, individuals who shift toward a more homosexual identity may have their first heterosexual activity at a later age because of their relative lack of sexual attraction toward the opposite sex, and those who move toward homosexuality may have an earlier age of first homosexual experience.

Homosexual identity formation is largely a cognitive activity. Individuals are not likely to place themselves in a category if they have little information about those who fit in that category or believe themselves to be dissimilar to those who are considered to be in the category. Therefore, prior to identifying as homosexual, a person must know homosexuality and homosexuals exist, learn something about what homosexuals are like as people, and be able to perceive similarities between their desires and behaviors and those of people labeled as homosexual (Troiden, 1989). Society's pervasive homophobia makes this progression more difficult. Consequently, the transitional process may require significant cognitive work and introspection. Hence, it is possible that individuals who function at a higher cognitive level may be more likely to move along the continuum or to move at a faster rate. Because education is thought to foster and to reflect cognitive ability and problem solving, individuals who are more highly educated may be more likely to move along the continuum as well. In support of this speculation, Prestage et al. (1995) found that among their Australian sample of men who had sex with men, those who self-identified as homo sexual reported higher levels of education than those who did not so identify. In addition, the characteristically liberal and protective environment of college may be the first place where they are able to meet self-identified homosexual peers; this exposure may facilitate movement toward homosexual self-identity. Finally, education may simply provide the language and/or perceived need to label one's feelings and behavior as "homosexuality." Thus, level of education may be associated with movement toward homosexuality.

In the current study we examined longitudinal data from a sample of behaviorally bisexual men. We predicted that a subset of respondents would move toward the homosexual and of the continuum of self-rated sexual orientation from Time 1 to Time 2 and that this group would be larger than the group of men who would move toward the heterosexual end of the continuum. We tested the following hypotheses about variables that may predict movement toward homosexual feelings. In each case we compared men who had reported moving toward more homosexual feelings (toward homosexual movers) with those who had not (comparison group).

Hypothesis 1. Toward homosexual movers will be more self-homophobic, anxious, and depressed and will report lower self esteem at Time 1.

Hypothesis 2. Toward homosexual movers will report more substance use in general and with their male sexual partners at Time 1.

Hypothesis 3. Toward homosexual movers will report more homosexual activity at Time 1. Specifically, they will be more likely to be in a relationship with a man, will be more likely to have ever had a steady relationship with a man, and will have had more male partners both within the past six months and in their lifetime. In addition, toward homosexual movers will have an earlier age of first homosexual activity.

Hypothesis 4. Toward homosexual movers will report less heterosexual activity at Time 1. Specifically, they will be less likely to be in a relationship with a woman, will be less likely to have ever had a steady relationship with a woman, and will have had fewer female partners both within the past six months and in their lifetime. We also predicted that the toward homosexual movers will have a later age of first heterosexual activity.

Hypothesis 5. Toward homosexual movers will be more likely to fantasize about men during masturbation at Time 1.

Hypothesis 6. Toward homosexual movers will be more highly educated.

Methods

Respondents

The original sample of 536 men were recruited during 1991 and 1992 as part of a study on behaviors related to the transmission of HIV. Because of the original focus on HIV, we recruited relatively young (18- to 30-year-old) men who had had penetrative sex with a man and a woman in the past three years. Penetrative sex was defined as oral or anal sex with a man and as vaginal, anal, or insertive oral sex with a woman. About half the original sample were Black, and about half were White. (See McKirnan, Stokes, Doll, & Burzette, 1995, for more details about recruitment procedures and characteristics of the original sample.) Recruitment sources included print advertisements that targeted "men who have sexual experience with both men and women" in both Black- and White-oriented newspapers (28% of the respondents); outreach in community settings, including coffee houses, street fairs, and public parks where men look or casual sexual partners (31%); outreach in gay and "mixed" (gay and heterosexual) bars (19%); and snowball sampling wherein respondents referred

Measures

Demographics. Measures included age, highest education level, income, and employment status. Income for the previous year was measured on a six-point scale ranging from (1) "less than $10,000" to (6) "over $60,000." Employment and educational status were asked as open-ended items and coded for standard categories.

Sexual orientation. Respondents were asked which point on a seven-point scale "best describes your current overall sexual feelings." The scale ranged from "completely homosexual" (1) through "equally homosexual and heterosexual" (4) to "completely heterosexual" (7), with each point labeled. Respondents also were asked which of three labels best described their current sexual feelings or sexual orientation: bisexual, gay, or straight.

Masturbation fantasy. Respondents were asked about the types of sexual partners they typically thought about when masturbating. Possible responses ranged from "almost always men" (1) through "about equally men and women" (3) to "almost always women" (5), with each point labeled.

Mental health. Four measures related to mental health were included. Self-homophobia, defined as negative attitudes toward oneself as a result of homosexual feelings and/or behaviors, was measured with a three-item scale ([Alpha] = .87; e.g., "Sometimes I wish I were not attracted to men"). Responses were rated on a five-point scale from disagree (1) to agree (5). The measures of anxiety and depression were six-item subscales from the Brief Symptom Inventory (Derogatis & Melisatatos, 1983) ([Alpha] = .81 for anxiety; e.g., "feeling tense or keyed up"; [Alpha] = .84 for depression; e.g., "feeling no interest in things"). Respondents indicated on a five-point scale the degree to which they had experienced each symptom during the past week. Self-esteem was measured with three items ([Alpha] = .54; e.g., "I wish I could have more respect for myself") obtained from an existing scale (Rosenberg, scale described earlier.

Substance use. The frequency with which respondents used alcohol or other drugs before or during sex with men over the previous six months was assessed. Respondents made separate ratings for alcohol use and drug use with each of three types of sexual partners: steady partner or lover, someone they knew well but who was not a steady sexual partner, and casual or anonymous partners. Responses were made on a four-point scale from "never" (0) to "always" (3) and summed across partner types. Men with no male sexual partners in the past six months (n = 21) have missing data on these variables. In addition, respondents gave information about general levels of alcohol and drug (marijuana and cocaine) use during the past six months, on a scale where 0 = never, 1 = rarely, 2 once a month or so, etc.

Gay identification and involvement in the gay community. Gay identification, the degree to which men consider their homosexual attraction important to their self-identity and their sense of being part of a gay community, was measured with four items in the disagree-agree format ([Alpha] = .66; e.g., "Being attracted to men is very important to my sense of who I am"). Outness, or disclosure of homosexual behavior to others, was the mean of nine items (a = .79). For each of nine people or groups (e.g., father, mother, siblings, close male friends), the respondent indicated whether he had discussed his homosexual behavior or feelings (rated 3), whether he thought the person/group suspected but had not been explicitly informed (2), or whether he thought the person/group had no knowledge (1). The mean of the nine ratings was the index of outness. Gay community involvement was measured with three items ([Alpha] = .62). Two items--the frequency with which the respondent read gay newspapers and the frequency with which he had contact with gay organizations--were rated on a five-point scale, ranging from "never" (1) to "several times a week or daily" (5). The third item was the respondents' number of gay friends, segmented into five ranges. Gay bar-going frequency in the past year was measured with a seven-point scale ranging from "never" (0) to "about every day" (6).

Number of sexual partners. Respondents estimated their number of male and female sexual partners for the previous six months and for their life.

Relationship status. We asked respondents if they were currently in a steady relationship with a woman, a man, or both, as well as how many steady relationships they had ever had with men and women. A steady relationship was defined as someone "who you are emotionally involved with and have had sex with."

Procedures

Potential respondents were given a university phone number to call for a screening interview. The phone number and times to call were included in advertisements and on hand cards distributed by outreach recruiters and previous study participants. Respondents who called were screened on the telephone with a brief questionnaire that included relevant demographic and behavioral items, as well as distractor items designed to conceal the actual screening criteria. An interview was scheduled with respondents who met the eligibility criteria.

Respondents were paid $25 for the initial one-on-one interviews. The initial interview was confidential, and respondents had the option of remaining anonymous, within the constraints of a face-to-face interview. At the end of the initial interview, however, respondents were invited to give us identifying information so that we could contact them again in about one year for a second interview. We successfully recontacted 216 respondents for a second face-to-face, confidential interview, which was conducted about one year after the first; these men are the sample for the current study. Respondents were paid $40 for the second interview.

Interview locations for both interviews included a university setting, the respondent's home, and interviewers' homes, depending on the preference of the respondent. Respondents were given their choice of gender and race of interviewer. All interviewers were trained and evaluated in basic interviewing and questioning skills, as well as on the specific interview schedule used in this study. When structured response scales were used during the interview, respondents were given a card containing the appropriate scale and asked to respond by number.

Results

Overview

After we compared the original and follow-up samples on relevant variables, we tested for changes from Time 1 to Time 2 in self-rated sexual orientation. We then tested specific hypotheses using multivariate analyses of variance (MANOVA), logistic regression, and t-tests as appropriate. MANOVA was used to test for group differences in sets of continuous variables. When the predictor variables included both dichotomous and continuous variables, we used logistic regression, which allows for regressing a dichotomous criterion variable on both continuous and dichotomous predictor variables. The significance test for the contribution of individual predictor variables in the logistic regressions was the Wald statistic, which has a Chi-square distribution with one degree of freedom (Norusis, 1990).

Because the distributions of number of sexual partners were extremely skewed, with some men reporting a very large number of partners, these variables were truncated at about the 98th percentile and transformed to their common logarithm. For ease of interpretation, reported means were computed from the untransformed, truncated variables, even when statistical tests were computed with transformed variables.

To test hypotheses about movement toward more homosexual feelings, two groups were defined. The first, toward homosexual movers (n = 73), consisted of men whose self-rated sexual orientation changed one point or more toward a more homosexual rating from Time 1 to Time 2. The second group, the comparison group (n = 101), included men whose self-rating of sexual orientation was the same at Times 1 and 2 (n = 64), plus men whose self-rating changed toward a more heterosexual rating from Time 1 to Time 2 (n = 37). It seems reasonable to include this latter group in the comparison group, because they clearly were not moving toward homosexuality. We omitted from these analyses the 42 men who did not move from Time 1 to Time 2 but whose self-rating at both time points was "completely homosexual" or "primarily homosexual." These men did not qualify for the toward homosexual movers group because they did not move from Time 1 to Time 2. It seemed illogical to include them 1 comparison group designed to provide a contrast to men who were moving toward a more homosexual orientation, because they were already homosexually identified.

To reduce the likelihood of Type I errors (falsely rejecting the null hypothesis), in looking for predictors of movement toward homosexuality, we tested only specific a priori hypotheses. In addition, when multiple dependent or predictor variables were involved, we used multivariate analyses (MANOVA or logistic regression) that allowed for a single statistical test for each hypothesis. The sets of dependent variables for MANOVAs were mental health (self-homophobia, depression, anxiety, self-esteem) and substance use (alcohol use and drug use in general and with male sexual partners). The sexual behavior and relationship variables were considered together in logistic regressions.

Comparing Men in the Follow-up Sample and Those Who Were Not Reached for Follow Up

There were two points at which respondents were lost to follow up. About one quarter (25.7%) of the original sample chose to remain anonymous and gave us no identifying information. We successfully recontacted 54.5% of those who gave us identifying information. Of those who had given us location information but were not re-interviewed, for most (86%) the address and phone number we had was not current; a few (9%) declined to be re-interviewed, 3% failed to return multiple phone calls, and three (2%) had died.

Respondents in the original and follow-up samples differed on e and education. The men in the follow-up sample were disproportionately White (58.8% of those followed up vs. 39.7% of those not reached for follow UP, [chi square][1] = 18.69,p [is less than].001) and college graduates (33.2% vs. 22.9%, [chi square] (2) = 11.09, p [is less than] .004). The two groups did not significantly differ on age (M = 25), income (Md = $10,000 to 20,000), or recruitment source.

The follow-up sample reported less sexual involvement with women than did the men who were not re-interviewed. Specifically, the men in the follow-up sample reported fewer lifetime female partners, Ms = 13.53 and 24.63, t(532) = -5.75, p [is less than] .001; fewer female partners in the past six months, Ms = 1.05 and 2.38, 0532) = -3.72, p [is less than] .001; and a later age of first heterosexual experience, Ms = 16.09 and 15.45, 0532) = 2.47, p [is less than] .01.

The men in the follow-up sample were less likely currently to be in a relationship with a woman (21% vs. 3 1%, [chi square] (1) = 7.16, p [is less than] .007) and more likely currently to be in a relationship with a man (32% vs. 24%, [chi square] (1) = 3.75, p [is less than] .05) than the men for whom we did not obtain follow-up data. Additionally, the men in the follow-up sample were more likely to have ever had a relationship with a man (80% VS. 70%, [chi square] (1) = 6.28, p [is less than] .01).

Respondents' three-point sexual orientation label was related to follow-up status, [chi square] (2) = 22.88, p [is less than].001; men who described themselves as gay were more likely to be included in the follow-up sample (55.8%) than were men who described themselves as bisexual (34.6%) or straight (28.3%). The follow-up sample was more gay identified on the self-rated seven-point sexual orientation scale, Ms = 3.13 and 3.8 1, t(53 1) = -5.56, p [is less than].001, and more likely to fantasize about men during masturbation, Ms = 2.12 and 2.63, 0452) = -4.31, p [is less than] .001.

Changes in Sexual Orientation from Initial to Follow-up Interviews

On the self-rated seven-point sexual orientation scale, 73 respondents moved toward homosexuality (34%), 37 moved toward heterosexuality (17%), and 106 did not change (49%) from Time 1 to Time 2. The respondents as a group had moved toward the homosexual end of self-rated sexual orientation, Ms = 3.13 (Time 1) and 2.91 (Time 2), t(215) = 3.15, p [is less than] .002. Of the respondents who moved, significantly more moved toward a homosexual orientation than toward a heterosexual orientation (676% vs. 34%, [chi square] (1) = 11.78, p [is less than] .001). In addition, at follow up, respondents reported a greater likelihood of fantasizing about men during masturbation, Ms = 2.13 (Time 1) and 1.98 (Time 2), t(173) = 2.52, p [is less than] .02, than they had at the initial interview.

Given that the men in the follow-up sample were more gay identified at Time 1 than the men we did not reach for follow up, respondents who were interviewed a second time may have already been in the process of moving toward homosexuality. Such men may have been more likely to continue to move toward homosexuality, making the previous percentage an overestimate of the number of men who were moving toward homosexuality. To counter this possible bias, we re-analyzed these data, restricting the sample to the men who identified as bisexual or straight at Time 1. This sample (n = 122), who did not differ on self-rated sexual orientation at Time 1 from the men we did not reach for follow up, Ms = 3.88 and 3.81, 043 7) = .49, p [is less than] .62, should provide a more conservative estimate of the percentage of behaviorally bisexual men who moved toward homosexuality. The percentages, however, were strikingly similar in this restricted sample: 32.8% moved toward homosexuality, 19.7% moved toward heterosexuality, and 47.5% did not change from Time 1 to Time 2.

Hypotheses Comparing Toward Homosexual Movers and the Comparison Group

To validate the single-item measure that was used to define the toward homosexual movers and the comparison group, we examined changes from Time 1 to Time 2 for these two groups on the following variables, which should reflect movement toward a more homosexual orientation: gay identification, outness, gay community involvement, gay bar-going frequency, and self-homophobia. We used a 2 (toward homosexual movers vs. comparison group) by 2 (Times 1 & 2) repeated measures MANOVA and predicted a significant interaction, such that the toward homosexual movers would show more change over time than the comparison group. The interaction was significant, mult. F(5,165) = 4.70, p [is less than] .001; means and univariate tests are shown in Table 1. For each individual variable except gay community involvement, movement from Time 1 to Time 2 in the direction consistent with a gay identity was greater for toward homosexual movers than for the comparison group.

Table 1

Means, Standard Deviations, and Univariate Tests of the
Interaction of Group and Time for Variables that Reflect
Movement toward Homosexuality

                        Toward Homosexual   Comparison
                               Movers         Group
                              (n = 73)      (n = 101)
Variable                  Time 1  Time 2  Time 1  Time 2  F(1,169)
Gay identification         3.31    3.42    3.04    2.86    5.33(*)
                           (.97)  (1.04)  (1.00)   (.93)
Outness                    2.07    2.31    1.86    1.95    6.90(**)
                           (.56)   (.53)   (.50)   (.54)
Gay community involvement  1.66    1.97    1.40    1.49    3.56
                           (.83)   (.80)   (.78)   (.80)
Gay bar-going frequency    2.92    3.13    2.44    2.08    6.38(*)
                          (1.68)  (1.59)  (1.63)  (1.48)
Self-homophobia            2.57    2.02    2.86    2.69    7.16(**)
                          (1.21)  (1.10)  (1.35)  (1.30)

Note: Standard deviations are in parentheses.

(*) p < .05; (**)p < .01

There were no statistically significant differences between the toward homosexual movers and the comparison group on any demographic variables. Comparisons on hypothesized differences between toward homosexual movers and the comparison group are described next.

Hypothesis 1: Mental health variables. There were significant differences between the toward homosexual movers and the comparison group on the set of mental health variables, mult. F(4,168) = 2.42, p [is less than] .05. Contrary to the original hypothesis, however, at Time 1, toward homosexual movers reported better mental health than did the comparison group. Univariate tests showed that toward homosexual movers were less depressed, Ms = .80 and 1.09, F(1,171) = 6.49, p [is less than] .02; less anxious, Ms = .88 and 1.22, F(1,171) = 8.07, p [is less than] .005; and had higher self-esteem, Ms = 4.07 and 3.76, F(1,171) = 4.62, p [is less than] .04, than did the comparison group. There was no significant univariate difference between groups on self-homophobia, F(171) = 1.90, p [is less than] .17.

Hypothesis 2: Substance use. Contrary to our hypothesis, toward homosexual movers reported significantly less substance use with their male partners at Time 1, mult. F(4,148) = 2.98, p [is less than] .03. Univariate tests revealed a toward homosexual movers were significantly less likely to use drugs with their male partners than the comparison group, Ms = .31 and .61, F(1,151) = 5.74, p [is less than] .02. There was no significant univariate difference regarding alcohol use with male partners between the groups, F(1,151) = .06, p [is less than] .81. Similarly, univariate tests revealed that the two groups differed on overall use of drugs, Ms = .78 and 1.38, F(1,151) = 7.73, p [is less than] .006, but not on overall use of alcohol, F(1,151) = .30, p [is less than] .59.

Hypothesis 3: Number of male sexual Partners and relationships with men. To test whether the measures of homosexual activity predicted movement toward more homosexual sexual feelings, we performed a logistic regression with this set of variables as predictors and group membership (toward homosexual movers vs. comparison group) as the criterion variable. Contrary to our hypothesis, the set of variables that reflected male sexual partners and relationships with men was not predictive of movement toward a more homosexual orientation, [chi square] (4) = 2.15, p [is less than] .71, nor were any individual variables different for the two groups. Moreover, the two groups did not differ on the age at which they had their first homosexual experience, Ms = 17.21 and 17.32, t(172) = -. 17, p [is less than] .87.

Hypothesis 4: Number of female sexual partners and relationships with women. Consistent with our hypothesis, variables that reflected lifetime and current heterosexual activity, as a set, predicted movement toward a more homosexual orientation, [chi square] (4) = 17.86, p [is less than] .002. Toward homosexual movers reported, at the initial interview, fewer lifetime female partners, Ms = 10.73 and 16.90, t(172) = -2.15, p [is less than] .04, and fewer female partners in the past six months, Ms = .84 and 1.56, 0172) = -3.16, p [is less than] .002, than did comparison group. There were no significant differences between the two groups in their likelihood of currently being in a relationship with a woman, 19.2% vs. 29.7%, [chi square] (1) = 2.48, p [is less than] . 12, or of their ever having been in a steady relationship with a woman, 90.4% vs. 97.0%, [chi square] (1) = 3.43, p [is less than] .07. Of the variables in this set, only the number of female partners in the past six months made a unique contribution to predicting group membership, Wald [chi square] [11 = 6.29, p [is less than].02. With homosexual experiences controlled, the set of variables reflecting heterosexual activity was still a significant predictor of movement toward more homosexual sexual feelings, [chi square] [41 = 17.86, p [is less than] .002. As hypothesized, toward homosexual movers reported a later age of first heterosexual experience, Ms = 16.49 and 15.32, t(172) = 2.66, p [is less than] .01.

Hypothesis 5: Masturbation fantasies. Toward homosexual movers reported a greater tendency to fantasize about men during masturbation, Ms = 2.09 and 2.50, t(154) = -2.09, p [is less than] .04, than did the comparison group.

Hypothesis 6: Education. There was no significant difference in educational level between toward homosexual movers and the comparison group, (Chi square) (2) = .65, p [is less than] .73.

Toward homosexual movers and the comparison group did not significantly differ on self-rated seven-point sexual orientation at Time 1, Ms = 3.47 and 3.51, respectively, t(172) = -.19, p [is less than] .85. However, movement was related to three-point sexual orientation label, (Chi square) 2 (2) = 11.81, p [is less than] .003. Toward homosexual movers were more likely than the comparison group to identify themselves as gay (42.3% vs. 18.4%) and less likely to identify themselves as bisexual (49.3% vs. 72.4%).

Discussion

Sample Bias

The men we reached for a follow-up interview differed in important ways from the men we did not reach. The men we reached were more likely to be White, college educated, and gay identified, and they reported fewer lifetime and current female sexual partners than the men we did not reach. Men who are more gay identified and more educated may be more comfortable disclosing same-sex activity and, therefore, more likely to provide follow-up information and to participate in a follow-up interview. They may also be more concerned about HIV and more interested in a study that has implications for HIV-risk reduction. College-educated men may be less transient and easier to find for follow-up interviews as well.

Because the follow-up sample was more gay identified at Time 1 than the group of men we did not reach, the follow-up sample may be biased toward respondents who were already in the process (or further along in the process) of moving toward homosexuality. Consequently, one could argue that the data from our follow-up sample may overestimate the percentage of behaviorally bisexual men who are moving toward more homosexual sexual feelings, although about the same proportions were obtained when the sample was restricted to men who initially identified as bisexual or heterosexual. On the other hand, the one-year time period between data-collection points is probably not adequate to capture changes in sexual orientation for all the men in our sample. For some men, the process of moving toward homosexuality (or toward heterosexuality) may cover a much longer time period. Hence, the proportion of men who might eventually move toward more homosexual feelings might be greater than the proportion in our sample. Ideally, we would use multiple follow-up interviews covering a longer period of time.

Changes in Sexual Orientation

At follow up, the sample as a whole had moved toward the homosexual side of self-rated sexual orientation. At Time 2, respondents were more likely to fantasize about men during masturbation than at Time 1. These results are consistent with the idea that at least some behaviorally bisexual men will, over time, become more exclusively homosexual in their desires. Although the sample as a whole made these changes, these group-level data do not reflect the varied experiences of the men involved. Only about one third of the sample moved toward more homosexual feelings in the course of a year. Half of the sample were stable in their self-rated sexual orientation from Time 1 to Time 2; clearly not everyone was moving toward a homosexual identity. Determining the percentage who will eventually self-identify as homosexual is problematic, especially given the impossibility of finding a representative sample of behaviorally bisexual men. In this sample, about one third moved toward more homosexual feelings in the course of a year. Given the bias in our follow-up sample toward more gay-identified men, this proportion may be an overestimate.

Predictors of Movement Toward More Homosexual Sexual Feelings

Mental health. Better mental health was predictor of movement toward the homosexual side of the continuum. Toward homosexual movers reported less depression, anxiety, and higher self-esteem at Time 1 than did the comparison group. This result contradicted our hypothesis, which had teen based on a previous finding that men who moved toward homosexuality had poorer mental health at the initial interview than did men not shifting toward homosexuality (Stokes et al., 1993). There were important sampling differences between the two studies. In the Stokes et al. (1993) study, self-identity as bisexual rather than bisexual behavior was the inclusion criterion, and there was no age restriction for respondents. The current sample was also larger and was recruited from a variety of sources, whereas the sample in Stokes et al. (1993) was found only through print advertisements. The samples differed also on self-rated sexual orientation. The current sample was more gay identified on a self-rated seven-point scale at both time points (Time 1: Ms = 3.13 and 4.00; Time 2: Ms = 2.90 and 3.60, for the current sample and Stokes et al., 1993, respectively). Thus, at Time 1, respondents in the current sample had already moved further along the sexual orientation continuum. Therefore, respondents in the current study may have been more comfortable with and accustomed to their same-sex attraction and behavior than those in the previous study.

In addition, the men who were reached for follow up in the Stokes et al. (1993) study were more self-homophobic than their original sample. Conversely, the men who consented to follow up in the current study were less self-homophobic than our original sample. Stokes et al. mentioned that the men in their follow-up sample seemed distressed about their same-sex activity and hypothesized that having a nonjudgmental person to talk to about issues of sexuality may have been therapeutic for more self-homophobic respondents, resulting in a higher rate of consent for follow up. Thus, their follow-up sample may have consisted of those men who were having the most difficulty integrating same-sex attraction into their lives and were therefore the most distressed. The toward homosexual movers in the current study seem to have been more comfortable with their same-sex attraction than were the men moving toward homosexuality in the Stokes et al. (1993) sample.

Substance use. Toward homosexual movers were less likely to use drugs other than alcohol both in general and with their male partners at Time 1 than men not moving in this direction. The two groups did not differ, however, on either measure of alcohol use. If the increased drug use for men in the comparison group relative to the toward homosexual movers were associated with increased distress about their homosexual activity, one would expect to find differences on alcohol use as well. Perhaps drug use, which was more common in the comparison group, was associated with experimenting with same-sex activity, or perhaps some men in the comparison group engaged in homosexual activity when the secondary gain of drug use was present.

Number of sexual partners and relationship status. The group of variables that reflected the number of female sexual partners and relationship history with women predicted movement toward more homosexual sexual feelings. Toward homosexual movers reported, at the initial interview, fewer lifetime female partners and fewer female partners in the past six months. A later age of first heterosexual experience was a significant predictor as well. Surprisingly, the number of male sexual partners, history of relationship with men, and age of first homosexual experience were not predictors of movement toward more homosexual feelings. For this sample, it was not the number of homosexual partners and relationships but the relative absence of heterosexual experiences that predicted movement toward the homosexual side of the continuum. The hypothesis that toward homosexual movers would be more likely than the comparison group to fantasize about men during masturbation was supported. Fantasizing about male partners during masturbation predicted movement, although sexual behavior with men did not.

The men who moved toward a more homosexual identity may be those who are not interested in having sex with women and have little choice but to shift. Conversely, if a man is able to enjoy heterosexual sex, he may not move toward homosexuality even if he is more active with male than female partners. Given society's prejudices against homosexuality, men who have heterosexual sex--even if their sexual involvement with women is infrequent--may be inclined to self-label as heterosexual (or bisexual). This speculation suggests a somewhat different conception of bisexuality than the one based on the scale introduced by Kinsey, Pomeroy, and Martin (1948), in which heterosexual and homosexual are opposite ends of a bipolar continuum, with bisexual somewhere in the middle. Storms (1978, 1980) suggested that homosexuality and heterosexuality be treated as separate dimensions, although this conception of sexual orientation has not been widely embraced. We agree with Storms and argue that attraction to men (AttM) and attraction to women (AttW) are relatively independent, unidimension scales. For males, being high on AttW and very low on AttM would lead to heterosexual feelings and behaviors. Similarly, one high on AttM and low on AttW would be homosexual. Bisexual feelings and/or behaviors would occur when one is above a certain treshold on both AttW and AttM. Using this model, our data suggest that AttW is more important in determining eventual sexual identity among behaviorally bisexual men. If one is above a threshold on AttW, one will likely not self-identify as homosexual, even if one's AttM is high.

Toward homosexual movers and the comparison group did not significantly differ on a self-rated seven-point sexual orientation scale at Time 1. The mean score for both groups was between "equally homosexual and heterosexual" and "more homosexual than heterosexual." However, there were differences between the groups on three-point sexual orientation labels, with toward homosexual movers being more likely to identify themselves as gay and less likely to identify themselves as bisexual than the comparison group. Thus, it appears that toward homosexual movers were more comfortable identifying themselves as gay than the comparison group and that this comfort level was associated with further movement toward homosexuality.

Summary and Conclusions

As previously noted, theorists have proposed that some individuals who self-identify as bisexual and/or who are behaviorally bisexual are actually in transition to another orientation (Gagnon, 1977; Klein, 1978; Stokes et al., 1993; Troiden, 1989). Others have hypothesized that adopting a bisexual identity may be less anxiety provoking than adopting a homosexual identity, which may more closely reflect the person's true sexual orientation (Coleman, 1987). Our data were consistent with this hypothesis. However, only about one third of the respondents in the follow-up sample had moved toward a more homosexual identity. Half of the sample did not shift in sexual orientation self-ratings from Time 1 to Time 2, and some respondents shifted toward the heterosexual side of the continuum.

Our data indicated that an absence of heterosexual experience, both current and lifetime, is a predictor of movement toward homosexuality within a sample of behaviorally bisexual men. Given the dissimilar results between our study and the Stokes et al. (1993) study, it appears that mental health as a predictor of movement is dependent upon where along the sexual orientation continuum the sample is when data are collected. That is, for more gay-identified samples, better mental health appears to facilitate further movement toward the homosexual side of the continuum. Less gay-identified samples will probably be less comfortable with their homosexual attraction. Movement toward homosexuality in such samples is likely to be associated with higher levels of distress.

The population of men who have sex with both men and women is complex and dynamic. Men vary considerably in how comfortable they are with their same-sex attraction and behavior. Self-identified sexual orientation does not always correspond well to actual sexual behavior, and self-identity may change over time (Rust, 1993). Of those who move toward homosexuality, some will eventually self-identify as gay; others will stop intermittently or indefinitely en route. Therefore, people working with behaviorally bisexual men would be wise to adopt a flexible bisexual, schema that allows for myriad variations in identification, behavior, and self-acceptance.

References

Cass, V. C. (1979). Homosexual identity formation: A theoretical model. Journal of Homosexuality, 4, 219-235.

Coleman, E. (1982). Developmental stages of the coming out process. Journal of Homosexuality, 7, 31-43.

Coleman, E. (1987). Bisexuality: Challenging our understanding of human sexuality and sexual orientation. In E. E. Shelp (Ed.), Sexuality and medicine (pp. 225-242). New York: D. Reidel.

Dank, B. M. (1971). Coming out in the gay world. Psychiatry, 34, 180-197.

Derogatis, L. R., & Melisatatos, N. (1983). The brief symptom inventory: An introductory report. Psychological Medicine, 13, 595-605.

Firestein, B. (Ed.). (1996). Bisexuality: The psychology and politics of an invisible minority. Thousand Oaks, CA: Sage.

Fox, R. C. (1995). Bisexual identities. In A. R. D'Augelli & C. J. Patterson (Eds.), Lesbian, gay, and bisexual identities over the lifespan (pp. 48-86). New York: Oxford University Press.

Friend, R. A. (1991). Older lesbian and gay people: A theory of successful aging. Journal of Homosexuality, 20, 99-118.

Gagnon, J. H. (1977). Human sexualities. Glenview, CA: Scott, Foresman.

Garber, M. (1995). Vice versa: Bisexuality and the eroticism of everyday life. New York: Simon & Schuster.

Kinsey, A., Pomeroy, W., & Martin, C. W. (1948). Sexual behavior in the human male. Philadelphia: W.B. Saunders.

Klein, F. (1978). The bisexual option. New York: Arbor House.

Leland, J. (1995, July 17). Bisexuality. Newsweek, 44-50.

Lever, J. (1994, August 23). Sexual revelations: The 1994 Advocate survey of sexuality and relationships: The men. The Advocate, 661 / 662, 16-24.

McKirnan, D. J., Stokes, J. P., Doll, L., & Burzette, R. G. (1995). Bisexually active men: Social characteristics and sexual behavior. The Journal of Sex Research, 32, 64-75.

Minton, H. L., & McDonald, G. J. (1984). Homosexual identity formation as a developmental process. Journal of Homosexuality, 9, 91-104.

Norusis, M. J./SPSS. (1990). SPSS/PC+ advanced statistics 4.0. Chicago: SPSS.

Peterson, P. L., Ostrow, D. G., & McKirnan, D.J. (1991). Behavioral interventions for the primary prevention of HIV infection among homosexual and bisexual men. Journal of Primary Prevention, 12, 19-34.

Prestage, G., Kippax, S., Crawford, J., Noble, J., Van de Ven, P., Campbell, D., Baxter, D., & Cooper, D. (1995). Report C.2: Sexual identity and sexual behaviour with both men and women in a sample of homosexually-active men in Sydney, Australia. Sydney: HIV/AIDS and Society.

Rosenberg, M. (1965). Society and the adolescent self image. Princeton, NJ: Princeton University Press.

Rust, P. C. (1993). "Coming out" in the age of social constructionism: Sexual identity formation among lesbian and bisexual women. Gender and Society, 7, 50-77.

Stokes, J., McKirnan, D., & Burzette, R. (1993). Sexual behavior, condom use, disclosure of sexuality, and stability of sexual orientation in bisexual men. The Journal of Sex Research, 30, 203-213.

Storms, M. D. (1978). Sexual orientation and self-perception. In P. Pliner, K. R. Blanstein, I. M. Spigel, T. Alloway, & L. Krames (Eds.), Advances in the study of communication and affect: Vol. 5. Perception of emotion in self and others (pp. 165-180). New York: Plenum.

Storms, M. D. (1980). Theories of sexual orientation. Journal of Personality and Social Psychology, 38, 783-792.

Troiden, R. R. (1979). Becoming homosexual: A model of gay identity acquisition. Psychiatry, 42, 362-373.

Troiden, R. R. (1989). The formation of homosexual identities. Journal of Homosexuality, 17,43-73.

Troiden R. R., & Goode, E. (1980). Variables related to the acquisition of a gay identity. Journal of Homosexuality, 5, 383-392.

Manuscript accepted January 13, 1997

Support for this research was provided through Cooperative Agreement Number U64/CCU506809-02 with the Centers for Disease Control and Prevention.

Requests for reprints and other correspondence about this article should be addressed to Joseph Stokes, Ph.D., Department of Psychology (m/c 285), University of Illinois at Chicago, 1007 W. Harrison St., Chicago, IL 60607. Phone: [phone]. Fax: [phone]. E-mail: [e-mail].