In the setting of a prepaid medical care plan, self-administered questionnaires were an accurate source of information on smoking habits and the standard against which the physiologic measures of smoking, serum thiocyanate, and expired carbon monoxide, must be judged. Questionnaire responses were internally consistent and highly reproducible. In contrast, the physiologic measures had low overall sensitivity as tests of current tobacco exposure (0.72). Their sensitivity was improved by utilizing the self-reported information.