The aim of this study was to document patterns of survival after resection for cure for gastric cancer in a well-defined population. A population-based series of 649 gastric cancers resected for cure between 1976 and 1995 in a 494000 population, was used. Resection for cure was performed in 44.4% of the diagnosed cases. This proportion increased from 36.8% (1976-1979) to 45.0% (1992-1995) (P=0.03) whilst operative mortality decreased from 18.3 to 12.7% (P=0.003). The overall crude 5-year survival rate (excluding operative mortality) was 32.6% (95% confidence interval (CI) 28.7-36. 5) and the corresponding relative survival rate was 40.9%. Prognosis did not improve during the study period. Stage at diagnosis was the most important prognostic factor, the 5-year relative survival rate being 81.2% (+/-5.9) in TNM stage IA, 76.9% (+/-8.0) in stage IB, 50. 4% (+/-4.6) in stage II, 24.4% (+/-3.7) in stage IIIA, 5.6% (+/-3.2) in stage IIIB and 5.2% (+/- 2.2) in stage IV. Stage at diagnosis, age, subsite and macroscopic type of growth were independent prognostic factors, in a multivariate relative survival model. Earlier detection or development of an effective adjuvant therapy could contribute to improvement in prognosis.