To evaluate retrospectively the clinical staging in a consecutive series of patients selected for cystectomy and to define its limitations with a view to possible improvements. From 1979 to 1988, 276 patients with newly detected or recurring transitional cell carcinoma (TCC) of the bladder, were offered pre-operative irradiation (20 Gy) and cystectomy. The patients were assessed during 1995 and the outcome related to both clinical and surgical data. Survival was analysed on the basis of 'intention to treat'. Estimates of survival probabilities were calculated by the method of Kaplan and Meier. Differences in survival among subgroups were assessed using the log rank test and Cox stepwise regression analysis. Cancer-specific actuarial survival for the whole series was 68% at 5 years and 63% at 10 years. Survival was closely related to the depth of invasion found at surgery, clearly discriminating those with tumours confined to the bladder wall (< or = P3A) from those with extravesical extension (> or = P3B). The cancer-specific survival at 5 years for patients with < or = P3A tumours was 85% and for those with > or = P3B tumours was 50%. This important distinction was anticipated accurately using bimanual palpation before surgery, those patients with no palpable mass after transurethral resection of bladder tumour (TURBT) having an actuarial survival of 83%, and those with a residual mass a survival of 50% at 5 years. In the multivariate analysis, increasing clinical stage was the only pretreatment variable with significant prognostic value for survival. However, this variable was highly dependent on the palpatory findings after TURBT, the presence of a residual mass being a prerequisite for the clinical stage T3 in case of muscle-invasive tumour. Bimanual palpation remains crucially important in clinical staging, and there is a need for further standardization and refinement of this procedure.