The results of two case-control studies of breast cancer which included questions on exposure to tobacco and alcohol are reported. One study included 998 hospital cases and a like number of matched hospital controls while the other included 118 cases identified during mammographic screening and a like number of matched normal screenees. Both studies used the same questionnaires and the same methods to obtain information. The results with regard to cigarette smoking differed between the two studies. The hospital-based study showed a decreased risk of breast cancer with increasing amounts smoked (relative risk for 15 or more cigarettes per day was 0.82, 95% confidence interval 0.60-1.13) while the screening study showed an increased risk (relative risk for 15 or more cigarettes per day was 2.90, 95% confidence interval 1.16-7.25). Evidence is presented that both results may be attributable to bias in the selection of cases and controls. It is concluded that reliable results on the relationship between smoking and breast cancer are only likely to come from population-based studies. These studies, in general, have found no relationship. Neither study produced any hint of an association between alcohol consumption and breast cancer. From this, it appears that bias in subject selection may not be such a significant factor in interpretation of studies of alcohol and breast cancer as it is in studies of smoking and the disease. A number of other difficulties in the interpretation of studies of alcohol and breast cancer are considered, including the great variation in the amount of alcohol consumed. It is concluded that the assertion that alcohol is a risk factor for breast cancer remains unproven.