Familial patterns of the occurrence of breast cancer were studied in a population-based case-control series of black women from the Cancer and Steroid Hormone study. The risks of breast cancer among relatives of breast cancer cases were compared to those of controls who were matched for age and locale. Using the term "proband" to indicate either case or control status, significant predictors of risk to the relatives of probands included case/control status of the proband and the number of years of education completed by the proband. Genetic segregation analysis of the case families using external risks generated from SEER data indicated that the familial aggregation was consistent with Mendelian recessive transmission of a single major gene. The use of internally estimated risks, which are much less stable than the SEER risks, no longer permitted discrimination among the major locus models examined. To avoid possible reporting bias, we also performed segregation analysis on families of probands who had completed at least 12 years of education. The results from this analysis reflected the results from the entire data.