It has been postulated that the relationship of environmental tobacco smoke (ETS) exposure to cancer or cardiovascular diseases may be confounded by social class or diet because women exposed to ETS by their smoker spouse belong to lower social classes and have an unhealthy diet. In a population survey in Geneva, Switzerland, 914 female never-smokers were interviewed about sociodemographic factors, health habits including a semiquantitative food frequency questionnaire, and exposure to ETS according to the site (home, work, leisure). Compared to women unexposed to ETS, those exposed to ETS at work ate less fibers, cereals, vegetables, lean meat, had a lower intake of iron and beta-carotene, and had a lower total energy intake; women exposed during leisure time ate less cereals, drank less skim milk, and had a lower intake of complex carbohydrates. But the diet of women exposed at home did not differ from the diet of those unexposed to ETS. Thus, "living with a smoker" in Geneva does not necessarily imply adopting his health and dietary habits. We conclude that confounding factors of the association of ETS and disease vary according to site and populations and therefore should not be invoked as a systematic source of bias in all studies.