As part of the National Institute of Mental Health Collaborative Program on the Psychobiology of Depression study, data were collected on 2225 first-degree relatives of 612 probands. We analyzed 187 families of bipolar patients (149 probands with a diagnosis of bipolar I disorder and 38 with a diagnosis of schizoaffective, manic subtype). Using traditional genetic methods, the morbid risk of bipolar illness in relatives was found to be 5.7% in the relatives of bipolar probands as contrasted with 1.1% in the relatives of probands with major depression. These values compared closely with those obtained using survival analysis. Relatives of probands with early onset were found to have a greater risk than relatives of probands with late onset. The sex of the relative, the sex of the proband, or the subtype of the proband (bipolar I or schizoaffective bipolar) did not influence the risk in the relative. The age at onset was found to be accelerated with birth cohort, with individuals born in more recent cohorts having an earlier onset. Multifactorial analysis found significant heterogeneity for sex-specific sibling correlations (with the brother-sister correlation smaller than the same-sexed correlations), and path analysis estimated transmissibility of liability to be 71%. The mixed model, which allows for a single major locus with a multifactorial background, gave evidence for the presence of a major locus when controlling for the effects of birth cohort and age at onset. However, this evidence is tempered when comparing the mixed model with a more general transmission model.