OBJECTIVE In the field of breast cancer control, obtaining population-based data on spontaneous mammography (MG) screening should be a priority. This study focuses on the breast cancer incidence and stage distribution in relation to MG use in Ravenna (Italy), 1987-88. METHODS We estimated the MG rates, expected (E) incidence based on mortality data, observed (O) incidence, predicted excess incidence based on MG rates, observed excess incidence, and stage distribution according to the MG history. RESULTS The highest MG rate (37%) was found among residents aged 40-44 but none of these had a T1a-bNO breast cancer diagnosed nor was the predicted excess incidence demonstrated. Between 45 and 64 years, 80% of self-selected screenees had repeat ("incidence") MG and the O:E incidence ratio was 1.32 (95% CI 1.09-1.58). The observed excess incidence was 3.8-fold (95% CI 2.56-5.16) greater than that predicted. Advanced (T2+ and/or N1+) cases accounted for 42% of patients diagnosed within 3 years of their last MG, for 55% of those diagnosed more than 3 years after their last MG, and for 70% of those with no previous MG. Above age 70, a significant worsening of stage at diagnosis was associated with a clearcut drop self-referral for MG. CONCLUSIONS The results indicate that self-selection and its implications are major features of spontaneous screening practice.