Evidence of the benefits of screening for breast cancer using mammography after the age of 50 is considered sufficient. Information on the determinants of compliance to the recommendations on mammography is needed to identify women to which public health messages should be specifically addressed and also to interpret results from epidemiological data in which some breast cancer cases are detected through screening programmes and others are not. The general characteristics and dietary data of French women participating in the E3N cohort study were analysed. Odds ratios of the frequency of non-compliance to recommendations on breast cancer screening were computed in women over and under the age of 50. Non-compliant women over 50 (i.e. never attenders) had a poorer access to physicians and poorer health with regard to specific risk factors even after adjustment for age and educational level. Women who were used to referring to gynaecologists early in life were better attenders later on. Logically, women under 50 who were over-screened, as compared to public health recommendations, had the opposite characteristics. Public health recommendations should be designed specifically according to targeted subgroups of women. Determinants of attendance for screening should be kept in mind in the interpretation of epidemiological studies in which some cases may be over-screened and others not.