Family history of cancer was evaluated for 79 breast cancer probands from among a series of consecutively ascertained cancer patients undergoing treatment in the authors' Oncology clinic. Cancer prevalence for each family was quantified by using a statistic that accounts for variable size and age structure among families. To test the null hypothesis that cancer risk is independent of family membership, the distribution of this statistic for families in their original configuration was compared with the distributions observed when relatives were randomly assigned to families in 99 random permutations of family membership. The results indicated significant heterogeneity for cancer risk among relatives of breast cancer probands, which suggests that the isolation of high risk families can provide a meaningful resource for in-depth studies in breast cancer genetics. Using an objective criterion for selection of such families, high risk for early onset breast cancer was confirmed in a prospective follow-up study fof four families from the original resource.