Despite the large number of surgical operations for stomach cancer in Japan, no study based on data from population-based cancer registries has been conducted regarding the relationship between hospital surgical volume for stomach cancer and patients' survival. Using data from the Osaka Cancer Registry (OCR), we performed survival analyses on 28,608 patients aged 35-79 years old who underwent surgery for stomach cancer, according to the extent of disease at diagnosis (localized: cancer is confined to the original organ; regional: cancer spreads to regional lymph nodes; adjacent: cancer infiltrates to adjacent tissue; distant: cancer metastasizes to distant organs). The study was divided into four periods; 1975-79, 1980-84, 1985-89 and 1990-94, according to year of diagnosis. Hospitals were stratified into four groups by total number of stomach cancer operations during each of the study periods, so that the numbers of operations in each group would be approximately equal. Cox's proportional hazards model was used to examine the relationships between hospital volume and 5-year survival according to the extent of disease, after adjustment for age at diagnosis and sex. Positive relationships between hospital volume and 5-year survival were seen in the localized and regional groups during the period 1975-79. These associations, however, decreased in later periods and disappeared in 1990-94 except for the very-low-volume hospitals. Stomach cancer patients treated in these hospitals remained at significantly higher risk of death than in the high-volume hospitals. In the case of the distant group, there was no clear relationship between hospital volume and 5-year survival during the study periods.