The effect of the month of diagnosis on survival was investigated in two series of unilateral invasive breast cancer, of which one comprised 95% of all such histologically diagnosed breast carcinomas in the city of Turku, Finland, in 1945 to 1965 (n = 401), and the other 94% of all such carcinomas diagnosed in 1980 to 1984 (n = 337). If the histological diagnosis was made in January, February, or August to October in 1945-65, or in July to September in 1980-84, mortality in breast cancer was greater than if the diagnosis was made during the rest of the year (P = 0.03 and 0.009, respectively). Cancers diagnosed during the unfavourable months had more tumour necrosis in both series, and higher mitotic count and larger tumour size in the 1945-65 series. The number of diagnosed cases was usually less than the median during the months associated with unfavourable prognosis. Hypotheses to explain the altering prognosis by the month of diagnosis include seasonal hormonal changes and social factors.