If it is correct to expand Kinsey's assumption to other aspects of sexual psychology, it should follow that whatever traits distinguish the HS and the HT orientations should also have intermediate values in bisexual (BI) individuals. Some of the attributes of interest might include gender atypical behavior in childhood, sex role, or sex-dimorphic traits in adulthood--and certain endocrinological variables which appear, in some studies, to significantly discriminate between HT and HS men (cf. Gladue et al., 1984). Although we do not mean to imply that such relationships would always be linear or even monotonic, bisexuality would be a questionable construct if it always turned out that self-identified bisexuals resembled self-identified homosexuals, for example, or heterosexuals. If average values of these traits, like sexual behavior and fantasy itself, lie for bisexuals on the continuum between corresponding values for homosexuals and heterosexuals, the Kinsey assumption would appear to have a powerful validation.
There are two other possibilities, however. The BI individuals may appear the same as either the HS or the HT groups on one or more traits. This outcome suggests that BIs are essentially more like the men at one end of the sexual orientation spectrum, but for seemingly adventitious reasons they have the capacity for a more diverse sexual arousal pattern. Another possibility is that BI individuals may score either higher or lower then HS or HT samples on measures of such traits. This outcome suggests that BIs differ in some qualitative way, other than in sexual behaviors or fantasies, from the other two groups.
One problem in investigating these issues is that there is no clear agreement about who should be included in a BI group. Should it be anyone who does not score 0 (indicating exclusive heterosexuality) or 6 (indicating exclusive homosexuality) on the Kinsey scale? Should it be those who score only in the middle range (scores of 2, 3, and 4)? Or should we allow members of the subject population to self-select membership in the BI group? We chose the last alternative, recruiting subjects who regard themselves, for whatever reasons, as BI and studying them in comparison with self-identified HS and HT subjects on a variety of dependent measures. We are thus investigating bisexual identity, as opposed to bisexual orientation.
We made this choice for several reasons. First, we plan future analyses of bisexual orientation in our sample. If they lead to different results than those we report here, then the discrepancy is an interesting fact which we will report. But we did not want to clutter our initial analyses with these complications. Second, one of the major instruments we used to assess sexual orientation and identity, the Klein Sexual Orientation Grid (KSOG; Klein, 1978), was standardized on a population broken down by sexual identity, not by sexual orientation (Klein et al., 1985). This breakdown is arguably the standard one in this field; we return to this point in the Discussion.
We chose to study personality, adjustment, and lipid levels for a variety of reasons. Certain personality differences--especially those relating to masculine and feminine roles--have been shown in many studies to distinguish HS and HT samples (on average). BI scores on these measures would thus reveal patterns of similarity and difference with each of the other two groups. The question of adjustment has been extensively studied in HS versus HT populations, and is now by and large resolved in favor of the view that there are few differences between these two groups (Gonsiorek, 1991). However, we decided to included a brief assessment of this topic for historical reasons, and because adjustment has rarely been studied in bisexual populations. Finally, whether there are any differences in adult sex hormone levels in HSs and HTs is a subject of much debate (e.g., Meyer-Bahlburg, 1977), but there is no doubt that men and women on average differ in their metabolism of fat, and there was a suggestion in an early study that HS men may on average differ in fat metabolism hormones. Accordingly, we aimed to administer several measures of personality and to measure the levels of lipids (which are, of course, targets of the fat metabolism process) that might discriminate among the three subject groups. All of the dependent variables selected have been used in previous correlational studies of HS and HT men--few of which have included BI individuals.
We note the following facts about lipids: Triglycerides are lipids (fatty substances) involved in fat metabolism and storage of fatty acids. Cholesterol is a lipid thought to contribute to heart disease and atherosclerosis; it also is a precursor to sex steroid hormones and adreno-cortical hormones. Low density lipoproteins (LDL) are a type of lipid thought to increase the risk of heart disease, whereas high density lipoproteins (HDL, or so-called "good cholesterol") are thought to reduce the risk of heart disease.
Evans (1972) found differences between HS (n = 44) and HT (n = 11) men in height, weight, muscularity, several blood serum lipid levels, and androgen metabolites (e.g., androsterone/etiocholanolone ratio). Although several later studies failed to corroborate certain of these differences, others have successfully replicated some of them (Margolese and Janiger, 1973; Friedman et al., 1977). In Evans's study, the HS men had less subcutaneous fat and smaller muscle/bone mass. The HS subjects also had lower triglycerides, phospholipids, cholesterol, and beta-lipoproteins.
Extensive literature exists comparing differences between HS and HT subjects' responses to sex-role and gender-typical behavioral measures (e.g., Carroll, 1978; Weiss and Dain, 1979; for a recent review, see Pillard, 1991). Very few attempts have been made, however, to include BI subjects in such studies to see whether they fall, as might be expected, between the HT and HS samples on such measures, or whether they are more like either the HS or the HT subject groups. MacDonald (1983) cautioned that "Despite the suspected large numbers of bisexually oriented people, there has been almost no research on bisexuality." More important, that author warns that much of the research on homosexuality is thoroughly confounded by the inclusion of large numbers of BIs as HSs--making such research difficult to interpret.
One study using solely psychological measures as a preliminary attempt to study the nature of bisexuality was reported by Zinik (1985). He found that his BI sample displayed the highest levels of what he termed "cognitive flexibility" (when compared with HTs and HSs) and found no differences between subject groups on measures of anxiety, depression, and hostility. Similar findings have been reported by others (Weinberg and Williams, 1974; Nurius, 1983), leading Zinik to support a model of bisexuality built on the individual's ability to be socially adjusted and adaptive when faced with new opportunities.
Unfortunately, the subject samples studied by these investigators suffer from strong inter- and intragroup differences in gender, age, and educational background. For example, the mean age of the BI sample in Zinik's (1985) study was 20 years older than the HT sample, and the BI subjects were drawn from active social clubs and support groups in the San Francisco area--a subject pool perhaps not representative of the general population of self-identified BI men and women who are not active in supportive social networks.
It was our intent to reexamine the possibility of significant differences between HS, HT, and BI men on a set of personality, adjustment, and lipid level measures. We wanted to recruit all three subject groups in a similar fashion, allowing for more heterogeneity with regard to age, educational, and socioeconomic backgrounds.
All subjects were recruited through advertisements in major Boston newspapers, in 1982-1983, offering a $10 payment for participation. One newspaper, published daily, had a general readership. The other paper was published weekly, with a general readership demographically skewed toward younger, well-educated readers and college students.
All subjects responded to ads that did not mention any aspect of sexual orientation, except for nine BI men recruited through advertisements that included the word "bisexual" (before "males") once the heterosexual groups had been filled. Bisexuals were defined as those who labeled themselves BI, regardless of their Kinsey scale ratings. The result was a sample consisting of 30 HS subjects, 31 HT, and 29 BI.
Subsequently, each subject completed the Bem Sex Role Inventory (Bem, 1981), Klein Sexual Orientation Grid (Klein, 1978, Weinrich et al., 1993), California Psychological Inventory Femininity (Fy) Scale (Gough, 1957), Thomas-Zander Ego Strength Scale (Zander and Thomas, 1960), and a "Health and Development Questionnaire" developed by the authors.
The Klein Sexual Orientation Grid is essentially an expanded Kinsey scale that covers ratings for past, present, and ideal on items such as sexual fantasy, friendships and emotional closeness. (Table II presents the major items constituting this scale; see Weinrich et al., 1993, for a factor analysis of this measure on this sample.)
The Thomas-Zander Ego Strength Scale is a true/false self-report questionnaire resulting in a single score which measures two aspects of ego strength. According to Robinson and Shaver (1973), these two aspects are described as follows: "First is a person's capacity to be self- directing and to translate intentions consistently into behavior. Second is the ability to control and discharge tension without disrupting other psychological processes." For example, one of the 27 test items is: "I never persist at things very long without giving up."
Childhood gender nonconformity or (in boys) childhood "sissiness" has been consistently demonstrated to be a correlate of adult homosexual orientation (for example, Bell et al., 1981; Green, 1987). Accordingly, our Health and Development Questionnaire included three questions about this phenomenon, as well as single questions about depression and the subjects' own opinion of their psychological adjustment. We cannot pretend that these areas were evaluated in depth, but we decided in this study to aim at breadth. If significant correlations were found, they could be more fully investigated in larger, later samples.
Sexual orientation was assessed in two ways. The subjects' self-label consisted of his response to the question, "Do you consider yourself to be: * Heterosexual * Bisexual * Homosexual." His Kinsey rating was determined by his answer to the question, "Circle the number which best describes your current sexual behavior and fantasy:"--followed by the 7 possible answers (0 to 6), with descriptions of each worded precisely as Kinsey et al. (1948) had worded them.
Data were entered into computers and analyzed using JMP version 2.06b from the SAS Institute.
TABULAR DATA OMITTED
TABULAR DATA OMITTED
Table III. Sexual Identity and Childhood FemininityTotal HT BI HS (N = 90) Avoided fights in childhood Total n No 15 5 4 24 Yes 16 24 26 66 Percentage No 48 17 13 27 Yes 52 83 87 73 Pearson Chi sq. = 11.52; p < 0.004 Was as boyish as other boys in childhood Total n No 5 16 17 38 Yes 26 13 13 52 Percentage No 16 55 57 27 Yes 84 45 43 73 Pearson Chi sq. = 13.2; p < 0.002 Considered a sissy in childhood Total n No 27 15 14 56 Yes 4 14 16 34 Percentage No 87 52 47 62 Yes 13 48 53 38 Pearson Chi sq. = 12.6; p < 0.002 The various measures of adult gender-typical behavior produced similar results. First, listed in part A of Table IV are both the scores on the California Psychological Inventory Fy scale obtained from our administration of this scale, as well as the normative results abstracted from the test manual (Gough, 1957). The HT respondents as measured by this scale were on average less feminine than the other two groups; there was no significant difference between the BI and the HS men.
Part B of Table IV presents data from the Bem Sex Role Inventory and illustrates that the HS and BI groups tended to respond not only in more feminine ways but also in less masculine ways than the HT group did as measured by these two scales. However, the pattern was less strong for Bem Masculinity, in which the HT versus BI comparison was only a trend (p < 0.07). As above, there were no differences between the HS and BI groups.
Part C of Table IV illustrates that the BI group on average did have significantly lower ego strength in comparison with the other two groups on the Thomas-Zander Ego Strength Scale. There was also a weak trend (p = 0.11) for the HS group to have higher ego strength than the HT group.
This same pattern was found in response to the questions, "Have you ever felt depressed and/or lonely for more than two weeks at a time?" and "If so, how long do these periods last and how frequently do they occur?" The BI group reported experiencing depression more than the other two groups reported (bottom half of Table V; p < 0.05). There may also be a tendency for this group to report longer periods of depression in absolute terms, although not when the sample is restricted to those who have had some depression.
Table V. Sexual Identity and Adjustment Total HT BI HS (N = 90) Do you consider yourself to be someone who is troubled and not well-adjusted? Total n Yes 6 12 2 20 No 25 17 28 70 Percentage Yes 19 41 7 22 No 81 59 93 78 Pearson Chi sq. = 10.51; p < 0.006 Ever depressed Total n Never 20 11 20 51 At least once 11 18 10 39 Percentage Never 65 38 67 57 At least once 35 62 33 43 Pearson Chi sq. = 6.145, p < 0.05
If it were to turn out that there were differences in the sample as a whole which disappeared when the obese outliers were removed, that could suggest that the original results were due to a higher level of pathological lipid-level syndromes operating in one of our sexual identity groups more than another. If, on the other hand, it were to turn out that differences in the samples emerge only when the obese are removed, that might suggest that a few individuals were clouding a real, underlying correlation by unduly increasing variance in the lipid level measures under study.
As it turned out, the 7 excluded subjects were evenly distributed across the sexual identity categories: 2 self-identified as HT, 2 as BI, and 3 as HS. The exclusion process happened to eliminate the individual with the lowest cholesterol score (75 mg/dl), who was separated from the next lowest score by nearly 1.5 standard deviations. It also eliminated the individual with the lowest HDL score. All other eliminated individuals were well within the normal range of the corresponding distributions. This pattern in itself suggests that if there is a relationship between sexual identity and any clear pathologies of lipid levels, such a relationship is too small or the pathologies occur too rarely to have been detectable in our sample.
There is also the possibility that differences in height and/or weight among the three groups might account for any lipoprotein differences we report below. However, such group differences did not reach statistical significance (by ANOVA, all p's exceeded 0.20, even with obese outliers excluded). Likewise, several multivariate analyses failed to demonstrate group differences in height or weight by sexual identity. However, for completeness, let us note that in a multiple logistic regression of sexual identity (trichotomized as HS, BI, and HT) as a function of height and weight (arguably reversing the expected direction of causation), weight had a significant effect, Wald Chi sq.(1) = 6.57, p approximately equal to 0.01, and the significance of height was a trend, Wald Chi sq.(1) = 3.45, p < 0.07, with heterosexuals being heavier for their height than homosexuals, and bisexuals intermediate. This association held for the sample with the obese outliers excluded, Chi sq.(2) = 6.74, p < 0.04, but not for the sample as a whole.
There are two further questions about subject exclusion that we address here. First, should we also have eliminated the three tallest subjects, especially since one of them is borderline lean (as opposed to obese)? Separate analyses show that excluding these three does not substantially affect our results. Second, should we have included the two marginally obese (marked by + in Fig. 1)? Including or omitting them does not substantially affect our results; we decided to eliminate the 7 rather than the 5 because the two + individuals are among the 6 of the 7 who are the highest in absolute weight (as well as relative weight for height) in the entire sample.
Because the obese individuals were distributed across the sexual orientation categories, it might seem odd that their removal would reveal sexual orientation differences. However, outlying individuals may be obese to different degrees, and they disproportionately add to the variance taken into account by the statistical test. Strictly speaking, an ANOVA should not be performed on a group whose distribution is clearly nonnormal. Appropriate removal of outliers makes the remaining distribution more normal, so that conditions of the test are more likely to be fulfilled. Resulting p's can thus attain significance--especially, as is the case here, when the uncorrected p's were at trend levels already.
Responses to the KSOG items are of specific interest in this regard. Answers, when grouped into the "past," "present," and "ideal" categories, seem to have little noticeable effect here (for a contrasting sample, see Weinrich et al., 1993). When answers to the KSOG are grouped by type of sexual orientation (horizontally rather than vertically), there is likewise little or no effect for three of the seven groups (the items concerning emotional closeness, socializing with, and leading the lifestyle of).
In contrast, the groups of KSOG items concerning sexual attractions, sexual fantasies, sexual relations, and sexual self-definition all show significant effects by sexual identity. On these items, HS, HT, and BI subjects gave answers that were significantly different from each other. However, the magnitudes of the differences are unequal: although the BI responses tend to be between the HS and HT responses, they are far closer to the HS side.
The BI group also differed significantly from the HS group on measures related to ego strength, depression, and self-esteem. For these variables, both the HT and HS groups scored significantly higher than the BI group. This finding is consistent with results obtained by Bursik (1981), utilizing the Thomas-Zander Ego Strength Scale, with smaller samples.
We also found similarities between the HS and BI groups in two aspects of lipid levels (high density lipoprotein and cholesterol), as well as one significant difference (triglycerides). Evans (1972) concluded from his study of physiological and physical differences that "that data support the thesis of an unidentified common factor underlying physical and personality characteristics of homosexuality." We do not address whether this common factor is environmentally controlled or results from a biological predisposition for HS.
Finally, we found that BI and HS individuals were similar on the average amounts of reported childhood gender nonconformity, but very different from HT men.
Let us now return to the question of whether our BI subjects were "really" bisexual, given their high scores on the Kinsey scale. Zinik (1985) established a set of three criteria that must be met to consider a subject BI. These criteria are (i) eroticizing both male and female in the form of sexual attraction and fantasies; (ii) having or desiring to have sex with both genders; and (iii) self-identification as BI as opposed to HS or HT. Wolf's (1985) sample, for example, was composed of BI men who had been (currently) married for over 2 years; all were sexually active within their marriages, and were also open about their homosexual behavior. The 73% of that sample identifying as BI clearly fits Zinik's criteria, even though most of the sample scored in the 4-6 ranges of the KSOG. (Note that the 4-6 range on the KSOG corresponds to the 3-5 range of the Kinsey Scale.) Our sample, then, is rather similar to the sample of BI used by this co-worker of Klein's.
Let us now turn to the sample used by Klein et al. (1985) in their original validation study of the KSOG. They wrote that "A canonical correlation analysis indicated that the vast majority of the variance (70%) between the two sets of variables sex, age, self-label vs. the 3 KSOG scale groups of past, present, and ideal^ was accounted for primarily by the respondent's self-identification". Moreover, one-third of the self- identified bisexuals did not fit into KSOG categories 3, 4, or 5 for the past and present time frames (2-4 on Kinsey Scale), although 77% of the bisexuals did fit into the middle range of the KSOG for the ideal time frame (only). Once again, we do not find that our BI sample is particularly unusual in its KSOG or Kinsey ratings when compared with other groups of self-identified BIs in the literature.
Bell et al. (1981) postulated two types of HS development, "predispositional homosexuality" and "learned homosexual responsiveness." Their findings suggest that exclusive homosexuality tends to "emerge from a deep-seated predisposition, while bisexuality is more subject to influence by social and sexual learning". Unsuccessful adaptation to such learning could result in self-identified BIs who are in conflict. These individuals constitute a category of sexual identity that is entirely different from either ego-syntonic homosexuals or BI men who have successfully adapted to their orientation, such as (possibly) men drawn from support groups and social clubs (the method of subject recruitment chosen by Zinik).
The closest we have come to an etiological model in this paper has been to examine childhood gender nonconformity in our three groups. A parsimonious explanation of our data in this regard suggests that childhood gender nonconformity is associated with an increased probability of same-sex attraction in adulthood (in comparison to HTs); that is, that HSs and BIs are rather similar to each other in the homosexual side of their sexual desires. What BI men have in addition is a higher recognition of their ability to be sexually attracted to women.
If, as we suspect, there are individuals who completely conform to the "flexibility" model of bisexuality--meaning successful social adaptation and constructive personal integration of their sexual identity--then we would suggest that the term "bisexuality," at least for males, actually refers to at least two very different groups of individuals. This is demonstrated by those subjects in our sample who conform to the "conflict" model of BI identification. That is, it appears that for these men, claims of bisexuality are employed as an "unconscious defense or a conscious denial" (Zinik, 1985). It is of interest that other dysfunctional psychological characteristics have been found to occur in higher proportions in BI subjects. Langevin et al. (1978) found that BIs scored significantly higher, on average, than exclusive HSs on the MMPI scale of paranoia, schizophrenia, and mania, whereas there were not between-group differences on the masculinity-femininity scale.
Upon careful review of the available data it seems evident that there may be a large population of "would-be homosexuals," who for various social and family reasons, are psychologically conflicted self-identified bisexuals. The term bisexual therefore seems to connote at least two separate groups of people with different patterns of sexual development, marked by either flexibility or conflict, who exhibit a common set of sexual behaviors.