It has been suggested that the well known associations between smoking and cancer may in part reflect inadequately controlled confounding due to occupational exposures. The purpose of the present analysis is to describe the association between cigarette smoking and both lung and bladder cancers, taking into account the potential confounding effects of over 300 covariates, most of which represent occupational exposures. A population-based case-control study was undertaken in Montreal to investigate the associations between a large variety of environmental and occupational exposures, on the one hand, and several types of cancer, on the other. Interviews were carried out with male incident cases of several sites of cancer, including 857 lung cancers and 484 bladder cancers. A group of non-smoking-related cancers, comprising 1,707 interviewed subjects, was used as one control group. Additionally, 533 population controls were interviewed and constituted a second control group. Interview information included detailed lifetime smoking histories, job histories, and other potential confounders. Each job history was reviewed by a team of experts who translated it into a history of occupational exposures. These occupational exposures, as well as nonoccupational covariates, were treated as potential confounders in the analysis of cigarette smoking effects. Regardless of whether population controls or cancer controls were used, the odds ratio (OR) between smoking and lung cancer (ranging from 12 to 16 for ever vs never smokers) was not materially affected by adjustment for occupational exposures. The odds ratios for bladder cancer (ranging from 2 to 3) were also unaffected by confounding due to occupational exposures. Because of the extensive information available on occupational exposures, it is possible to affirm that the reported associations between smoking and both lung and bladder cancer do not result from inadequate control for occupational confounders. The consistency of findings when using two different control groups further diminishes the possibility that the results represent an artifact of inadequate research design.