are consistent with a genetic influence hypotheses. Even so, concordance estimates for sexual orientation vary widely. Hershberger (2001), for example, reports data from 8 twin studies, with concordance rates between 0% and 100 % for sexual orientation for MZ twins. In most cases, concordance for DZ twins is reported to be lower than for MZ twins, except for King and McDonald (1992), and Hershberger (1997) for males. In more recent studies, which work with larger samples usually drawn from twin registries, concordance between twin pairs, and differences in concordance rates between MZ and DZ twins, are substantially lower than reported in earlier literature (Pillard and Bailey 1998, Hershberger 2001). For example, in 1952 one study reported 100 % concordance on sexual orientation for 37 pairs of MZ twins and 15% concordance among 29 pairs of DZ twins (Kallmann 1952a,b). In contrast, Kendler et al. (2000) report 31% concordance for sexual orientation for MZ twins and 13 % for DZ twins with data from a national probability sample of twins in the US. Other recent studies with samples drawn

from twin registries show concordances of 20%-25% for MZ twin pairs (Hershberger 2001). As samples become more representative, concordance on sexual behavior, attraction, and orientation, as expected, declines. Concordance is not always considered. Other researchers working with these same data do not report concordance rates but instead report estimates of heritability. Here, (narrow) heritability (h2) is defined as the ratio of additive genetic variance over total phenotypic variance. Kirk et al.

(2000) calculate heritability for sexual orientation at 50-60 % for women and 31 % for men. In contrast, Pillard and Bailey (1998) find zero heritability for women. Hershberger (1997) uses data from the Minnesota twin registry, which show no heritability for men but substantial heritability for women. Thus, heritability estimates for sexual orientation reported in the literature also vary widely. This inconsistency of results makes inference basically impossible. About the only finding that many researchers, including  social scientists (Peplau et al. 1994, Whisman 1996), agree on is that female homosexuality follows a different pattern than male homosexuality. It is unclear, however, what this pattern looks like, i.e. whether female sexuality is more (or less) .biological..

The problems with measuring heritability are substantial.5 It was originally conceived to compare the effects of selective breeding with environmental modification in agricultural experiments. Outside an experimental context, separating additive genetic variance of a trait from non-additive variance is difficult, if not impossible (McGuire 1995). Furthermore, differences between MZ twins and DZ twins in the impact of shared environments on behavioral outcomes may inflate estimates of heritability6. Consequently, behavior genetic models are more likely to overestimate

_________________________________________________

5 McGuire (1995) argues that heritability estimates are strictly valid only for the specific conditions under which they were derived. Specifically, phenotypic variance depends as much on the environment as on genes, more precisely, it is produced by gene-environment interaction. This, in addition to small sample sizes, may explain the wide variation in h2 estimates across samples, times, places. Incidentally, heritability estimates have no relationship to the nature versus nature question. For example, an instinct, which by definition is genetically determined, would show zero heritability (no trait variance in the population). Furthermore, h2 does not tell us anything about the etiology of a trait. One early twin study of prevalence of tuberculosis showed, for example, a correlation of 87.3 for MZ twins and 30.2 for DZ twins, which could be interpretable as a sign for high heritability (McGuire 1995). Yet we know that TB is caused by bacteria, and that environmental factors play a large role in its epidemiology, although obviously, genetic

predisposition to environmental factors could play a significant role in disease acquisition.

6 To pick just one example, the friendship networks of MZ and DZ twins are remarkably different; with MZ twins evidencing significantly greater overlap than same-sex DZ twins, especially with respect to alters who consider them as friends. Since adolescent behavior is associated with peer group structure, even subtle differences in friendship networks, not typically considered in behavior-genetic models, will have a significant impact on estimates of heritability.

Strana 4

than underestimate heritability. This problem is compounded by small samples and reliance on largely inadequate statistical methods (Jaccard and Dodge n.d.).7

Equally problematic, no twin study of sexual orientation except for Kendler et al. (2000) has, to our knowledge, worked with a probability sample. All early studies were based on clinical samples, convenience samples, or prisoners and other captive populations that are clearly biased.

Even for the twin registry studies, which avoid selecting on the dependent variable, biases are well known. MZ twins are much more likely to participate in twin studies than DZ twins (McGuire 1995, Lykken, McGue and Tellegen 1987), and males are more likely to enroll than are females (Hershberger 2001). Kendler and Eaves (19xx) report that twins who are more alike tend to volunteer for twin studies. Finally, participants in surveys about sexuality may be more educated, have more liberal attitudes, be more novelty-seeking, and experience earlier sexual debut (Dunne et al. 1997) than eligible non-participants. In contrast, our respondents, drawn from the National Longitudinal Study of Adolescent Health (Add Health) show no evidence of bias across a wide array of characteristics that may be associated with sexual behavior.

Even more problematic, data on pair concordance is most often derived from reports of only one person. One available test of the accuracy of such reports casts doubt on the validity of measures based on indirect reports. There is a less than 50 % chance that heterosexual twins will know that their co-twin is not heterosexual. More importantly, non-heterosexual persons are more likely than others to misidentify their heterosexual siblings as homosexual. This is also true for twins who were .absolutely certain. of the sexual orientation of their co-twin (Kirk, Bailey, and Martin 1999). In contrast, we consider data on attraction from direct self-report of each individual in the

sibling pair.

Potentially stronger support for the hypothesis that there is genetic influence on romantic samesex preferences come from studies (Hamer et al. 1993; Hu et al. 1995) which purport to provide evidence from molecular analysis of the X chromosome of male relatives of male homosexuals for an X-linked gene at position Xq28 associated with homosexuality. Recent work by Rice et al. (1999), however, suggests that there is little foundation for the Xq28 linkage hypothesis. Specifically, they find no support for the presence of a gene influencing sexual orientation at Xq28. This suggests that if there is a gene for sexual orientation, it is elsewhere on the chromosome. Considering all of the previous evidence for genetic influence on sexual orientation, one should be cautious in reaching the conclusion that there are such effects. Evidence from social surveys is often contaminated by strong selection effects and biological studies have failed

to identify a genetic marker for homosexuality. Given the striking cross-cultural variation in erotic preference, genetic expression, if present, must be very strongly conditioned by the sociocultural environment.

Evolutionary Dynamics

As noted above, if concordance rates do not parallel degree of genetic similarity, a simple genetic influence model should be rejected. Net of empirical evidence, many observers are troubled by the idea that simple evolutionary dynamics ought to limit the role that genetics could play in shaping same-sex attraction. Simply put, homosexuals are less likely to have children than others, and this simple fact ought to lead to a rejection of genetic determination of sexual orientation. The critique of genetic influence on this basis is relatively weak, and easily handled within an

__________________________________________________

7 Using an established method in behavior genetics, the DeFries-Fulker model, Jaccard and Dodge (n.d). calculate substantial heritability for caring for tropical fish (28 %), and frequency of various behaviors such as purchasing folk music in the past year (46 %), chewing gum (58 %), and riding a taxi (38 %).

Strana 5

evolutionary framework. Miller (2000), for example, posits that homosexuality may be a

.polygenetic. trait, that is, a trait influenced by a number of different genes, which, individually, result in greater fitness, and, only collectively result in homosexual orientation. Specifically, the idea is that these genes shift male brain development in a .female direction,. resulting in .greater sensitivity, tendermindedness, kindness, empathy. and therefore, .better fathers. as well.. Thus, the greater reproductive success of men whose genotype includes some of these genes, and the adverse effect on the reproductive success of men with all of them, cancel each other out, leading

to an evolutionary equilibrium that allows for homosexuality. This model suggests a link between gender identity and sexual attraction. At first glance, research findings showing a strong correlation of childhood gender-nonconformity and same-sex attraction

lend credence to this theory (Bell, Weinberg, and Hammersmith 1982; Dunne et al. 2000; Bailey and Zucker 1995). Among others, Bem (2000) suggests that childhood gender non-conformity represents the .developmental pathway. for the genetic expression of homosexual orientation. The evidence for this connection, though, is quite weak. The few prospective studies in this area focus on small clinical samples of extremely feminine boys, many of whom were diagnosed with gender-identity disorder (Green 1987). The vast majority of girls who display gender-atypical behavior grow up to become heterosexuals (Peplau et al. 1999).

Retrospective assessment of childhood behavior, the method that most studies use, is deeply problematic and likely to lead to overestimating the association between childhood behavior and adult identity simple because of the demands of narrative (Ross 1980; Bearman and Stovel 2000).

The association between childhood gender-atypical behavior and adult homosexuality, in this view, are created at the individual level in the form of life stories that have to make sense in the context of a culture that insists on equating gender and sexual identity.8

A second evolutionary theory about fitness and sexual orientation hypothesizes that homosexual orientation may increase .fitness. if it prevents later-born sons of large sibships to engage in unproductive competition with their older siblings (Miller 2000). The literature suggests some support for this idea, on first glance. Specifically, a relationship between birth order, or, more precisely, number of older brothers, and sexual orientation of males has been reported in a series of papers (Blanchard 1997; Blanchard and Bogaert 1996a,b; Purcell, Blanchard, and Zucker 2000; Bogaert 2000). No such effect was found for females. But the evidence and mechanism

proposed are extremely weak. These studies work with non-representative samples, and/or indirect reports on siblings. sexual orientation and suffer from the same biases as noted above in considering the genetic influence literature. Furthermore, the mechanism by which such an effect is thought to be activated seems somewhat far-fetched. Specifically, mothers are hypothesized to carry a .biological memory. (in the form of a H-Y Antigen) of how many sons they have carried, which leads to changes in the intra-uterine environment that activate .feminization. of younger sons (Blanchard and Klassen 1997, Miller 2000).

In this article, we test the second evolutionary model directly and find no support for an

association between birth-order and same-sex attraction. The first model, the idea that

homosexuality is a polygenetic trait cannot be tested with our data. Nevertheless, we show that concordance rates do not correspond to the general genetic model, and this fact alone falsifies the idea that there could be genetic influence in the absence of a social structural interaction.

_______________________________________________________

8 Riesman and Schwartz (1988) speculate that the observed decline in the proportion of lesbians who assume male roles and identities (.butch.) may be associated with the advent of an alternative narrative of identity for lesbians, namely, feminism.

Strana 6

Hormonal influences on sexual orientation

A number of researchers have proposed that same-sex preferences may be driven by hormonal imbalances resulting from exchange of hormones in utero. The logical chain involved is thin. The basic argument is that in rodents, sex hormones have been shown to transfer between fetuses in utero resulting in the expression of sexually dimorphic traits (Boklage 1985). This finding has given rise to the idea that opposite sex human twins will be affected in utero by the transfer of their siblings. hormones (Miller 1998, 1994; Dempsey et al. 1999; McFadden 1993; Rodgers et al. 1998). Specifically, at mid-term pregnancy, amniotic fluid shows large differences in testosterone levels between male and female fetuses. Since hormones are thought to cross the placenta and enter mothers. blood, a transfer of testosterone from a male twin to his twin sister in

utero is possible, leading to a .masculinization. of females. No reverse effect (.feminization. of males) is expected, as male and female fetuses do not differ with respect to the level of .female. hormones such as estrogen or progesterone (Miller 1998).9

Working through the argument, and starting with the first element, we find that the evidence for hormone transfer in humans is, at best, weak. Dempsey, Townsend, and Richards (1999) report that OS female twins have larger dental crowns (a male trait) than either SS female twins or singletons, whereas OS male twins dental crowns are not different than SS male twins or singletons. Males and females emit noises out of their ears. These noises, which we do not hear, are called spontaneous otoacoustic emissions (SOAEs) report that OS female twins emit half the average of SOAEs as SS female twins or singletons, suggesting that uterine exposure to androgens has masculinized their auditory systems (McFadden 1993). Both studies suggest some

.masculinization. of females, but not .feminization. of males, as expected.

With respect to more obviously social behaviors, gender stereotyped toy play, sensation seeking, and responses to public opinion questionnaires, the support for the intrauterine transfer hypothesis is weak (Rodgers et al 1998). Henderson and Berenbaum (1997) report no differences between OS twins and SS twins among 7-12 year olds in play behavior with gendered or neutral stereotyped toys. Miller (1994) reports that OS female twins age 3-8 play behavior did not differ from that of female SS twins. As with Resnick et al (1993) who report increased sensation seeking (a male trait) among female OS twins, but no .feminizing. effect for male OS twins, all of these studies are based on small-N convenience samples.10

No reliable evidence from human twin studies has shown intrauterine hormone transfer effects on males. Considering the second step in the argument, it is not exactly clear how such hormonal transfers would express themselves with respect to sexual preference.11 While some male homosexuals exhibit hyper-feminine traits, many male homosexuals exhibit hyper-masculine traits. Masculinity, in this context, is not a singularly heterosexual characteristic. Likewise, even if females were .masculinized. by androgen washing in utero, it is not clear why this would lead

________________________________________________

9 Huston (1983) describes findings from a number of studies exploring the effect on children of high doses of

progesteron or estrogen given to mothers with difficult pregnancies. Compared to control groups, either no effect was shown, or the differences between exposed and unexposed children did not follow the predicted pattern of, say, a propensity for feminine behavior, skills, or personality in boys.

10 Loehlin and Martin (2000) examine three variables that usually show gender differences (being worried, being reserved, and rule-breaking) for a large sample of twins from the Australian twin registry. The authors conclude that hormonal effects may too small to detect for even large samples; that previous obtained results, if any, may reflect postnatal socialization effects or may be due to sample fluctuation or measurement error.

11 The idea that prenatal exposure to sex hormones is associated with sexual behavior is derived from experiments with rats and guinea pigs which show hormone-induced sex-atypical behavior. For a critical review of the literature which interprets these findings as a socialization effect, see Fausto-Sterling (1995). A critical view on the comparison of rodents and humans with respect to sexual behavior and .orientation. is also found in Byne (1995), among others.

Strana 7