Multivariate structural equation modeling techniques have been applied to examine the causes of individual differences in responses to several items concerning sexual orientation. To minimize potential ascertainment and response biases, the study sample involved a large (N = 4901) community-based cohort of Australian twins aged 18-52 who answered an anonymous questionnaire on sexual behavior and attitudes. The statistical power of the analysis was increased by the availability of multiple measures of sexual orientation (behaviors, attitudes and feelings), providing stronger evidence for the existence of additive genetic influences on this phenotype than in a previous analysis (Bailey et al., 2000). Estimates of the heritability of homosexuality in this sample ranged between 50 and 60% in females but were significantly lower (heritability of approximately 30%) in males.