Cancer mortality among 40,498 Hiroshima and Nagasaki residents was examined in relation to cigarette smoking habits and estimated atomic bomb radiation exposure level. Relative risk (RR) models that are either multiplicative or additive in the two exposures were emphasized. Most analyses were directed toward all nonhematologic (ANH) cancer, stomach cancer, lung cancer, or digestive tract cancer other than stomach cancer, for which there were, respectively, 1,725, 658, 281, and 338 deaths in the follow-up period for this study. Persons heavily exposed to both cigarette smoke and radiation were found to have significantly lower cancer mortality than multiplicative RR models would suggest for ANH cancer, stomach cancer, and digestive tract cancer other than stomach cancer. Surprisingly, the RR function appeared not only to be submultiplicative for some of these cancer site categories but also may be subadditive. The lung cancer RR function could not be distinguished from either a multiplicative or an additive form. The number of deaths was sufficient to permit some more detailed study of ANH cancer mortality: RR functions appeared to be consistent between males and females, though a paucity of heavy smoking females limits the precision of this comparison. The submultiplicative nature of the RR function mentioned above was particularly pronounced among persons who were relatively young (less than or equal to 30 yr of age) at the time of radiation exposure. The RR function for these younger subjects depends strongly on both radiation and cigarette smoke exposure levels. Even light smoking (approximately 5 cigarettes/day) for an extended period of time was associated with a large estimated RR. Implications of these findings are discussed in relation to human carcinogenesis models. As a byproduct, cancer mortality of several sites is significantly related to radiation exposure in this population, after accommodation for the possible confounding effects of cigarette smoking.