The possible confounding caused by smoking was studied in connection with a mortality and cancer incidence study of 3,520 workers in the pulp and paper industry. A group 1,290 sawmill workers was used for comparison in addition to the expected numbers based on national statistics. A total of 801 questionnaires was sent to a representative sample of the workers, 537 to people still alive and 264 to the next-of-kin of decedents. The reply percentage was 86.6 for the former and 80.9 for the latter. The prevalence of smoking was determined for 1956 and 1981. The proportions of moderate and heavy smokers were assessed for 1981, and smoking indices were calculated for each occupational group. The estimated rate ratios for lung cancer in relation to smoking categories and the corresponding smoking-adjusted standardized incidence ratios were calculated for 1981, and they were compared with the observed standardized incidence ratios for lung cancer. The smoking habits could not explain the observed excess of lung cancer, nor the increased mortality from coronary heart disease found among the pulp and paper workers. Postal questionnaires may be a feasible tool for assessing smoking habits in retrospective cohort studies.