If the risk of disease is not the same for all germline mutations in a given gene, or if there are other familial modifiers of risk in carriers, then family-history-based estimates of average risk for detected mutations in that gene will depend on how carriers are sampled. Risk may also depend on the site or type of mutation. We studied 51 families with strong histories of breast cancer who attended Australian family cancer clinics and in which a germline mutation in BRCA1 or BRCA2 had been identified (28 and 23 families, respectively). Breast cancer risk in carriers was estimated under maximum likelihood theory, using information from all family members including those not tested, with adjustment for ascertainment by conditioning on genotype of the proband and family phenotype. The average cumulative risk of breast cancer for mutations in either BRCA1 or BRCA2 was 27% (95% confidence interval 16-43%) to age 50 and 64% (44-83%) to age 70. When grouped, the incidence in carriers was on average 17 (10-30) times that in non-carriers, independent of gene or mutation type (hazard ratios: 11 (4-29) for BRCA1, 23 (12-43) for BRCA2 (P for difference = 0.23); 13 (6-29) for protein-truncating mutations, 30 (9-104) for missense mutations and 30 (10-90) for splice-site mutations). For missense mutations, this was equivalent to a cumulative risk to age 70 of 83% (40-100%) and was due in part, but not totally, to the missense mutations 300 T>G in BRCA1 and 4486 G>T in BRCA2, which were individually found to be associated with high risk (P<0.001). Mutations in the central region of BRCA1 may be associated with a lower risk. The issue of the pathogenicity of specific variants may be addressed analytically providing there are one or more suitably informative families with that mutation.