Mortality from cancer and other causes in male and female first-degree relatives of women with breast cancer diagnosed before age 60 has been examined in a large population-based cohort study, providing estimates of familial risks free from ascertainment or recall bias. Relatives of 3,295 patients with breast cancer diagnosed in the UK between 1954 and 1981 were identified through a register of households established in 1939. The 11,678 first-degree relatives thus identified were followed up through national records until the end of 1992. Over this period 5,421 deaths (including 1,527 cancer deaths) occurred in these relatives. Mortality from breast cancer was significantly raised in first-degree relatives (SMR 187, 248 deaths), and there was also significant excess mortality from cancers of the larynx (SMR 177, 17 deaths), endometrium (SMR 166, 29 deaths) and unspecified neoplasms (SMR 153, 70 deaths). The SMR for ovarian cancer was 130, based on 58 deaths (p = 0.06). There was no marked excess for other sites or for non-neoplastic causes of death, but there was a significant deficit in mortality from cervical cancer (SMR 63, 18 deaths). The SMR for breast cancer increased significantly with decreasing age of the relative. After allowing for age, sisters of cases had a slightly (though non-significantly) higher risk than mothers (ratio of SMRs 1.22). These results, together with penetrance estimates from linked families, suggest that approximately one woman in 800 carries BRCAI, the susceptibility gene on chromosome 17q, and that this gene causes about 1% of all breast cancers.