In a case-control study of 1,004 lung cancer cases and 1,004 population controls, all nonsmokers (71 cases, 236 controls) were interviewed for their exposure to environmental tobacco smoke. On the basis of indices of duration, we separated intermediate and high exposures from low or no exposure. The odds ratio for high exposure was 2.09 (95% confidence interval = 1.02-4.28). Exposure from the spouse was only weakly associated with lung cancer risk. We found little association for exposure during childhood. High exposure at the work place showed an odds ratio of 1.91. There was little indication of confounding by dietary habits.