The relationship between smoking habits and occupational exposures was reviewed with regard to the effect of missing smoking information on occupational studies. A mathematical model for the assessment of the magnitude of confounding effects by smoking was described. The potential confounding effect of smoking determined by the model was relatively weak in relation to other occupational health factors. In general, the effects of not controlling for smoking were not significant unless the smoking habits of the base population were extreme. Methods of indirect control for smoking included analyzing other smoking related diseases in the study population, using an internal nonexposed cohort, using an adjustment factor based on hypothesized differences in smoking habits between the exposed and nonexposed groups, and a dose response analysis of the data. The effects of confounding due to smoking in case referent studies was discussed with regard to the representative nature of the referent population. A wide range of risk ratios attributable to sampling variation occasionally resulted in improper interpretation of the confounding variable, particularly in small reference populations. Other topics presented included the potential interactive effects of smoking and exposures (multiplicative versus additive models), the interpretation of studies lacking smoking data, and recommendations and needs for future methodologic research.