A prospective study was conducted to investigate the possible effect of breast self-examination (BSE) on cause-of-death-specific survival rate of breast cancer patients. Six hundred and four breast cancer patients diagnosed in 1984-1986 in Finland, and applying for breast prostheses, were interviewed about both their BSE practices prior to cancer diagnosis and the actual method of tumor detection. No clear differences were observed in the stage distribution or cause of death-specific five-year survival rates between individuals with different BSE practices. After adjustment for potential confounders in the Cox proportional hazards analysis, no differences in risk of breast cancer death were observed for those who performed BSE monthly as compared to those who practised BSE less frequently or not at all. When the method of detection was taken into account, it turned out that only 34 (7.6 %) of the 448 regular BSE practisers had actually detected their cancers by means of BSE. Furthermore, no survival advantage was associated with detection of breast cancer by means of BSE. Those BSE practisers whose cancer was detected by BSE had a similar or slightly worse prognosis compared to BSE practisers whose cancer had been detected by other means. Our results suggest that BSE practice is not beneficial in terms of breast cancer survival, nor is detection of breast cancer by means of BSE. Conclusive evidence should, however, be obtained from prospective randomized studies of breast cancer mortality.