A retrospective analysis has been made of a consecutive series of 3924 cases of breast carcinoma in women seen in one centre subject to a consistent management policy between 1954 and 1964. Survival was studied in relation to the clinical stages of the disease as defined by three different staging systems, Manchester, International TNM and that of the International Union Against Cancer and the American Joint Committee on Cancer Staging and End Results Reporting (UICC/AJC). In spite of considerable differences in the distribution of cases between stages according to the system applied there was markedly little difference in survival rates up to 20 years when each stage was compared with the other two. The most important prognostic factor used in staging is the size of the primary tumour. Although the frequency of clinical evidence of spread outside the breast is related to tumour size, significantly higher 20-year survival rates were found for patients in whom the tumour appeared to be confined to the breast on clinical examination when compared with patients in whom there was clinical evidence of spread beyond the breast only for tumours less than 5 cm maximum diameter. These results suggest that staging systems relying wholly on restricted clinical criteria are of limited value in terms of prognosis and their overfrequent modification is not justified.