Common, but weakly penetrant, functional polymorphisms probably account for most of the genetic risk for breast cancer in the general population. Current polygenic risk models assume that component genes act independently. To test for potential gene-gene interactions, single nucleotide polymorphisms in ten genes with known or predicted roles in breast carcinogenesis were examined in a case-control study of 631 Caucasian women diagnosed with breast cancer under the age of 53 years and 1,504 controls under the age of 53 years. Association of breast cancer risk with individual genes and with two- and three-gene combinations was analyzed. Sixty-nine oligogenotypes from 37 distinct two- and three-gene combinations met stringent criteria for significance. Significant odds ratios (ORs) covered a 12-fold range: 0.5-5.9. Of the observed ORs, 17% differed significantly from the ORs predicted by a model of independent gene action, suggesting epistasis, i.e., that these genes interact to affect breast cancer risk in a manner not predictable from single gene effects. Exploration of the biological basis for these oligogenic interactions might reveal etiologic or therapeutic insights into breast cancer and other cancers.