A prospective study was made on 3827 Japanese patients who had undergone partial gastrectomy for benign gastroduodenal diseases to examine whether they are at a high risk of mortality from primary gastric stump cancer (PGSC) and whether the risk is determined by the surgical procedure. The patients were followed up from the time of surgery (from 1948 to 1970) to June 30, 1981. Of 3,701 patients (96.7%), the vital status at the end of observation was determined, the total person-years at risk being 62,286.33. The observed deaths were compared with the expected deaths calculated from the mortality rates of Japan. An elapsed time of 10 years from operation to death was set not only to exclude possible recurrent, remaining, or multiple cancers but also to allow a certain latency period for the development of PGSC. The observed and expected deaths from PGSC were 11 and 52.85, respectively, the ratio being 0.21 (p less than 0.01). The ratios were uniformly less than 1 for both sexes and across three operative groups: Billroth I, Billroth II with Braun's anastomosis; or Billroth II without Braun's anastomosis. No difference was observed between the death rates from PGSC by operation type. The possible role of the postoperative nonphysiological (pathological) environment or duodenogastric reflux in gastric stump carcinogenesis was not detected in the present study.