Segregation analysis of breast cancer in families can provide the logical basis and the specific genetic models for mapping and identifying genes responsible for human breast cancer. Patterns of breast cancer occurrence in families were investigated by complex segregation analysis. In a sample of 1579 nuclear families ascertained through a population-based series of probands, an autosomal dominant model with a highly penetrant susceptibility allele fully explained disease clustering. From the maximum-likelihood Mendelian model, the frequency of the susceptibility allele was 0.0006 in the general population, and lifetime risk of breast cancer was 0.82 among susceptible women and 0.08 among women without the susceptibility allele. Inherited susceptibility affected only 4% of families in the sample: multiple cases of this relatively common disease occurred in other families by chance. The same genetic models, with higher gene frequency, explained disease clustering in an extended kindred at high risk of breast cancer. Evidence for a highly penetrant, autosomal dominant susceptibility allele for breast cancer in a high-risk family and the general population suggests that high-risk families can serve as models for understanding breast cancer in the population as a whole.