Stomach cancer mortality was prospectively studied among 9753 Japanese men and women who first responded to a mailed questionnaire in 1985 and were then followed through May 31, 1991. During this follow-up period, 57 stomach cancer deaths were identified. Current smokers had an increased risk of deaths from stomach cancer compared with never smokers (relative risk (RR) = 2.29, 95% confidence interval (CI): 1.15-4.56), but there was no dose-response to amount of cigarettes smoked. Daily alcohol drinkers who consumed 50 ml or more of alcohol per day also had a greater risk than nondrinkers (RR = 3.05, 95% CI: 1.35-6.91). There was no association between stomach cancer mortality and individual food consumption except a positive association with fruit intake. However, frequent use (greater than or equal to 3-4/week) of broiling of meats and traditional style Japanese salad preparation in their cooking procedures were positively associated with stomach cancer mortality. The RR values compared with infrequent use (less than or equal to 1-2/month) were 2.27 (95% CI: 1.06-4.85) and 3.10 (95% CI: 1.40-6.85), respectively. A positive family history of cancer, especially stomach cancer, significantly increased the risk of stomach cancer deaths (RR = 2.01, 95% CI: 1.12-3.63). The effects of these variables remained after adjustment for other variables.