Genetic Influence

We now test whether genetic influence on sexual orientation is expressed. Here, we use the data in its dyadic form. If genetic influence were expressed in these data, MZ twins should have the highest concordance for same-sex erotic preference, and unrelated and half-siblings the lowest.

Table 5 is based on pairs in which at least one respondent reports a same-sex romantic attraction (N=527 pairs).

Table 5 shows that there is no evidence for strong genetic influence on same-sex preference in this sample. Among MZ twins, 6.7 % are concordant. DZ twin pairs are 7.2% concordant. Fullsiblings are 5.5 % concordant. Clearly, the observed concordance rates do not correspond to degrees of genetic similarity. None of the comparisons between MZ twins and others in table 5 are even remotely significant17. If same-sex romantic attraction has a genetic component, it is massively overwhelmed by other factors. As argued above, it is more likely that any genetic influence, if present, can only be expressed in specific and circumscribed social structures.

_______________________________________________

 16 Table 4 shows odds ratios and associated 95% confidence intervals from a logistic regression with population weights. Standard errors are corrected for the sample design. The sample for table 4 is restricted to non-twins for whom self-reported total number of full siblings corresponds to the number of full siblings living in the household. Twins were excluded to avoid confounding with the opposite-sex twin effect reported above. Repeating the same analysis shown in table 4 for the full sample or various subsets, and with different operationalizations of sibship structure, such as the various indices specified in the literature (Blanchard 1997), did not yield a birth order effect.

17 Nevertheless, there is evidence of familial similarity across all pairs of related siblings . the probability that any randomly matched pair would be concordant under independence is less than 1%. Note also that for males (but not for females) the differences are in the expected direction (concordance rates increase with increasing relatedness). The percentage point differences are so small, however, that we would require a very large sample of twins to obtain statistical significance. The sample size needed for a two-sample comparison with the proportions estimated for male MZ and DZ twins from Add Health is approximately 795 twin pairs for each group. If prevalence and concordance rates reported in this paper approximate the true values, this requires approximately an 8 % sample of the entire relevant population in the US, a sampling strategy that may assure statistical significance for some social or psychological factors as well. Neither of these factors, however, is likely to play an important role in the etiology of same-sex attraction.

Strana 12

The single social structure we observe that is consistent with an argument for genetic expression is that of restricted gender socialization associated with first-born OS twin pairs.

Discussion

The findings presented in this paper confirm some findings from previous research and stand in marked contrast to most previous research in a number of respects. First, we find no evidence for intrauterine transfer of hormone effects on social behavior. Second, we find no support for genetic influences on same-sex preference net of social structural constraints. Third, we find no evidence for a speculative evolutionary model of homosexual preference. Finally, we find substantial indirect evidence in support of a socialization model at the individual level. Here we consider why our results differ from previous work. Subsequently, we consider the significance of these results for understanding the etiology of same-sex attraction.

Substantially higher concordance for homosexual orientation has been reported in previous research. We believe that previous work is largely incorrect as a result of reliance on nonrepresentative samples, for example, readers of gay publications, and reliance on indirect evidence. Specifically, while some studies obtained reports on sexual orientation from both siblings, others relied on one individual.s report on his or her sibling.s sexual orientation. These data structures are clearly associated with potential bias on the dependent variable. Kendler et al. (2000), however, report substantially higher concordance rates for self-reported sexual orientation

among adults in a study that overcomes some of these obvious methodological flaws. In this instance, the inflation of concordance may be a product of an interaction between small sample size and subtle selection dynamics. Specifically, their sibling and twin response rates were low.18

If individuals jointly participate in a study, and self-selection dynamics are present, as they likely are in this case, then concordance on traits other than willingness to participate in a study is to be expected. Consequently, we consider their concordance rates for same-sex orientation to be higher than would be expected under study designs less susceptible to self-selection.

In this study, we consider adolescent same-sex romantic attraction. The proportion of adolescents reporting same-sex attraction is significantly higher than the proportion reporting same-sex sexual experience. While it is possible that genetic expression on attraction is weak, whereas genetic expression on behavior is stronger, it seems more likely that the obverse should be true. Much of what we know about the etiology of adult homosexuality is derived from life stories of selfidentified homosexuals. These narratives often identify early same-sex romantic attraction as a constituent element in identity-formation. In addition to attraction, opportunity has to present itself. Since opportunity is clearly socially structured, our expectation is that social influences should be stronger for behavior than attraction.Whether a strong pathway between adolescent same-sex romantic attraction and self-identified homosexual identity exists, or whether it is the product of narrative demands for coherent lifestories, is unclear. There is clearly a strong association in our data between attraction and behavior, but the number of adolescents involved in homosexual relationships is too small in our sample to assess genetic influence statistically with any confidence. However, if the previous hypothesis were correct, it would suggest that socialization experiences might shape desire, but

____________________________________________________

18 Kendler et al. (2000) do not report the final response rate for their samples. From their sample description (p. 1844) we calculate a response rate of 18% for twin pairs, and 14% for the sample of siblings. The difference in concordance between MZ and DZ twins is not statistically significant (p=.203, own calculations based on data given in Kendler et al., table 1, p. 1845). In fact, inference about whether the proportion of concordant pairs among a population of MZ twins is .32 rather than .13 (the concordance reported for DZ twins) requires a sample size of at least 51 pairs, while Kendler et al. have data for only 19 pairs.

Strana 13

pathway from attraction to behavior. This study shows that for OS twins, in the absence of strong gender socialization, the proportion of male adolescents with same-sex attraction is twice as high as observed in the population as a whole. If there is genetic influence on same sex romantic preference, it expresses itself within a narrow and circumscribed social context characterized by equality. But this is exactly where one

would expect such expression, where social and cultural constraints governing sexual identity and orientation are least developed, and consequently, least constraining. Our findings reject simple genetic influence models. They are entirely consistent with a more general model that identifies the specific social structural contexts in which one would expect to observe genetic influence, for this, and an array of other outcome variables.

Social scientists not long ago left individual-level causes of sexual attraction and behavior to the biologists, choosing instead to focus on the striking cross-cultural differences in the organization of sexual expression. This article considers how such individual variation that we do observe could be organized. We test an old, and simple idea: culturally gender-neutral socialization experiences are likely to be associated with less patterned (for that culture) expressions of gender identity, of which sexual attraction is a key element. We find support for this idea: only in families with OS twins without an older same-sex sibling do we observe a substantially increased probability of same-sex attraction for males. For females, the observed rate in these contexts is roughly one-half the national norm. It is possible that some other subtle unmeasured dynamic is

going on, but as we can rule out simple genetic, hormonal, or evolutionary arguments, the main emphasis must point to socialization experiences. Here, we identify just one structure for socialization effects.

In general, social scientists hostile to the idea of genetic influence on social behavior should keep in mind the simple truism that without opportunity, genetic expression on behavior is impossible. Some stark examples should suffice: genetic expression for alcoholism is impossible in cultures without alcohol, population groups without food cannot express a genetic predisposition for obesity. Of course, examples of the complete elimination of opportunity for genetic expression are few and far between. Social structure may eliminate the possibility of genetic expression for some groups, but not all. This fact alone suggests one, perhaps paradoxical, reason why we observe an effect for male, but not female, OS twins. Against this background, therefore, the scope conditions of the findings reported in this article are also relatively clear. If there are no

main effects for genetics, we would not expect to observe genetic expression on romantic attraction except in cultures, like ours, where socialization regimes insist on the close linkage between cultural ideals of masculinity and femininity and sexual expression. In contrast, if there are genetic main effects, they would visible redominantly in cultures where such linkage is absent. Nonetheless, the idea that genetic influence, if present, should be insensitive to social categories in its expression is simply wrong. One should look to social structure to understand observed outcomes . especially for those that are thought to be shaped in some way by genetic inheritance.