Recent studies have identified a number of genes in the human genome at which germinal mutations predispose the individuals to one or another type of cancer. These studies also show that not all individuals carrying the mutant genes develop cancers (i.e., the mutant genes are not fully penetrant). At least some of these predisposed genotypes also have a higher sensitivity to cancers induced by ionizing radiation than those who are not so predisposed, which may be dependent on dose. This paper presents an analysis of the impact of such heterogeneity on estimates of cancer risks for an irradiated population. This is done by extending the Mendelian one-locus, two-allele model of cancer predisposition and radiosensitivity developed earlier to allow for incomplete penetrance and dose dependence of radiosensitivity differentials among genotypes. The model is applied to recently published data for breast cancer and hereditary non-polyposis colon cancer using a range of possible values for the strength of predisposition and radiosensitivity differentials. It is shown that, after radiation exposures, the ratio of cancer risks in a heterogeneous population relative to that in a homogeneous population increases with increasing dose, but that the dose dependence of the relative risk diminishes at higher doses. Likewise, the attributable risk (i.e. the proportion of the increase in risk that is due to both increased susceptibility and increased radiosensitivity) and the proportion of attributable risk due to increased radiosensitivity also increase with dose, and the dose dependence of each measurement also diminishes at higher doses. However, when the proportion of cancers due to the susceptible genotypes is small (<10%) (as is likely to be the case for breast cancer in non-Ashkenazi women), the increases in the relative risk and attributable risk are marked only when there are very large increases in cancer susceptibility (>1000-fold) and radiosensitivity (>100-fold) in the susceptible group. When the proportion of cancers due to the susceptible genotypes is appreciable (> or = 10%) (as may be the case for breast cancer in Ashkenazi Jewish women), there may be large increases in the relative risk and attributable risk for comparatively modest increases in cancer susceptibility (>10-fold) and radiosensitivity (>100-fold) in the susceptible subpopulation. For any given combination of strength of predisposition and radiosensitivity differential, incomplete penetrance dilutes the effect.