A family history of breast cancer is well established as a risk factor for the disease. Because family history is a dynamic rather than a static characteristic, longitudinal studies of entire families can be very instructive in quantifying the significance of risk classification. The Minnesota Breast Cancer Family Study is a historical cohort study of relatives of a consecutive series of 426 breast cancer cases (probands) identified between 1944 and 1952. The incidence of cancer and the measurement of risk factors in sisters, daughters, granddaughters, nieces, and marry-ins was determined through telephone interviews and mailed questionnaires. Ninety-eight percent of eligible families were recruited, and 93% of members participated. A total of 9073 at-risk women were studied: 56% were biological relatives of the case probands, whereas the others were related through marriage. Through 1996, 564 breast cancers were identified in nonprobands. Compared to the rate of breast cancer among marry-ins (188 cases), sisters and daughters of the probands were at a 1.9-fold greater age-adjusted risk (128 cases; 95% confidence interval, 1.4-2.4); granddaughters and nieces were at a 1.5-fold greater risk (248 cases, 95% confidence interval, 1.2-1.8). The breast cancer risk since 1952 was not distributed equally across families: although all biological relatives had a family history of breast cancer, 166 families (39%) experienced no additional cases. Most of the cases occurred among a subset of families: 21 families had 5 breast or ovarian cancers, 8 had 6, 2 had 7, and 4 had > or =8. There was no evidence of significantly increased risk for cancer at other sites, including the ovaries, cervix, uterus, colon, pancreas, stomach, or lymphatic tissue, although there was some evidence that stomach cancer in previous generations may help define the susceptible subset. These families contain four to five generations of validated occurrences of cancer, thus minimizing the uncertainty of genetic risk inherent in a disease with a late and variable age at onset. The patterns of breast cancer in these multigeneration families is consistent with the influence of autosomal dominant susceptibility in a subset, low penetrance genes in another, and purely environmental influences in the remainder.