Lung cancer risk estimates for exposure to environmental tobacco smoke remain controversial, a major unresolved issue being misclassification of smokers. We studied misclassification rates in two large cohorts using information on smoking obtained several years apart. Cohort I included Swedish twins born between 1886 and 1925 who answered questionnaires in 1961 and again in 1967 or 1970. Cohort II was a random stratified population sample of individuals born between 1894 and 1945 who responded to postal smoking surveys in 1963 and 1969. We considered those who stated that they had never smoked in the second questionnaire, but who reported smoking or former smoking in the first questionnaire, to be misclassified. In cohort I, 4.9% of male and 4.5% of female ever-smokers were misclassified, corresponding to 11.1% and 1.3% of reported never-smokers, respectively. Cohort II yielded similar results. A follow-up through 1992 of cohort I showed a relative risk for lung cancer among misclassified men of 1.9 [95% confidence interval (CI) = 0.4-9.1], as compared with 4.5 (95% CI = 2.0-9.9) and 13.3 (95% CI = 6.5-27.0) for former and current smokers, respectively. No case occurred among misclassified women. Although misclassification of smokers exists, our results indicate that it mainly concerns light smokers or long-time ex-smokers, who have only a very moderately elevated risk of lung cancer. It therefore appears unlikely that confounding by smoking explains the increased risk for lung cancer related to environmental tobacco smoke exposure.