A family-history cancer survey was conducted on 5,486 men who underwent a radical prostatectomy, for clinically localized prostate cancer, in the Department of Urology at the Mayo Clinic during 1966-95; 4,288 men responded to the survey. Complex segregation analysis was performed to assess the genetic basis of age at diagnosis and the familial clustering of prostate cancer. For the total group, no single-gene model of inheritance clearly explained familial clustering of disease, which could be partly explained by lack of Hardy-Weinberg equilibrium, with an excess of homozygotes. After accounting for deviations from Hardy-Weinberg equilibrium, the best-fitting model that explained the familial aggregation and age at diagnosis was a rare autosomal dominant susceptibility gene, and this model fitted best when probands were diagnosed at <60 years of age. The model predicts that the frequency of the susceptibility gene in the population is .006 and that the risk of prostate cancer by age 85 years is 89% among carriers of the gene and 3% among noncarriers. A strength of our study is its large size, such that genetic models could be fitted within strata defined by the age of the proband. Although the autosomal dominant model was consistently the best model, the parameter estimates differed somewhat (P=.03) across the different age groups, suggesting genetic heterogeneity. Additional evidence that the hereditary basis of prostate cancer is likely to be genetically complex was provided by the following: (1) there was a significantly elevated age-adjusted risk of prostate cancer among brothers of probands, compared with their fathers (relative risk 1.5 [95% confidence interval 1.4-1.7]); (2) the autosomal dominant model predicted an excess of homozygotes, over that predicted by Hardy-Weinberg equilibrium; and (3) the model-predicted risk of prostate cancer among relatives was inadequate when probands were diagnosed at age >=70 years.