Background. Systematic population-based screening for gastric cancer is widely spread in Japan. However, the case-control study method has been the main method used to evaluate the effectiveness of the screening to reduce gastric cancer mortality in Japan. Methods. This article presents a population-based cohort study. A questionnaire about lifestyles and dietary habits was distributed to 36,990 residents in a city of Japan. The response rate to the questionnaire was 92.0%. After ineligible responders had been excluded, 24,134 subjects were classified into screened and unscreened groups according to their self-reports of participation in the screening the previous year. We followed them up for 40 months and linked resident death records in the city. We compared mortality from gastric cancer and all other causes between the groups by using the Cox proportional hazard model. Results. The follow-up period was 78,156.6 person-years from September 1992 to December 1995. The multivariate relative risks for gastric cancer death of the screened group in comparison with the unscreened group were 0.72 (95% CI 0.31-1.66) among males and 1.46 (95% CI 0.43-4.90) among females. Conclusion. Although our data are preliminary, we were unable to demonstrate a large contribution of the present screening program to decreasing gastric cancer mortality. Further follow-up is needed to increase the precision.