Familial clustering of breast cancer has been recognised for over a century but until recently a genetic basis has been suspected rather than proven. Epidemiological studies have tended to support the view that an autosomal dominant gene, with high but incomplete penetrance, accounts for most breast cancer families. However, it is likely that several different predisposing genes are present within most populations. Difficulties arise in a conventional 'linkage mapping' approach to identifying these genes, first, because it is not clear that genetically homogeneous groups of families can be recognised on the basis, for example, of mean age of onset or pattern of other cancers within the kindred and, second, because breast cancer is so common (affecting almost one in twelve women) that large affected kindreds are likely to include an admixture of sporadic (non-genetic) cases. Cytogenetic and 'Loss of Heterozygosity' (LOH) studies in sporadic breast cancers have pointed to several candidate loci for breast cancer genes but there is no clear consensus from these two approaches that might direct attention to any prime target region. Recent reports of tight linkage between familial breast cancer (early onset) and breast/ovarian cancer (regardless of mean age of onset) and a locus on chromosome 17q21 defined by the anonymous probe CMM86, have not been confirmed in detail but have led to the identification of a locus some 15 Mb centromeric of CMM86 that gives a high positive lod at very low recombination fraction in fifteen Edinburgh breast and breast/ovarian cancer families. The disease in the majority of such families therefore appears to be attributable to a mutant gene at 17q12-21. A much smaller proportion of familial breast cancer is accounted for by mutations in the p53 gene (17p13). Not all such families fulfil the criteria for Li-Fraumeni syndrome and not all of the inherited mutations lie within exon 7 of p53. Counselling of members of breast cancer families becomes more exacting as these genetic lesions are identified. It is essential to extend the collection of data and tissue (blood or fixed pathology material) as widely as possible to confirm linkage to a specific locus within each individual kindred, to define the precise mutation and to establish the cancer phenotype and its penetrance. In the course of these studies a substantial population of women at high risk of breast (and other) cancer will be identified. Resources should be directed to this population so that optimum procedures for screening and prevention can be developed.