The relationship between alcohol drinking and gastric cancer risk was analyzed using data from a case-control study conducted in Northern Italy between 1985 and 1993 on 746 cases of histologically confirmed incident stomach cancer and 2,053 controls in hospital for acute nonneoplastic nondigestive tract diseases. Wine was the most frequently consumed alcoholic beverage, accounting for approximately 90% of all alcohol consumption. Compared with those who never drank wine, the odds ratios (OR) were 1.1 [95% confidence interval (CI) 0.9-1.3] for fewer than four drinks per day, 1.3 (95% CI 1.0-1.7) for four to fewer than six drinks per day, 1.6 (95% CI 1.1-2.4) for six to fewer than eight drinks per day, and 1.4 (95% CI 1.0-2.0) for eight or more drinks per day. No association was observed with beer or spirits. For total alcohol consumption, 25% of cases and 30% of controls never drank alcohol, and the multivariate OR for those who drank versus those who did not drink was 1.1 (95% CI 0.9-1.4). After allowance for smoking, education, family history of stomach cancer, selected micronutrient intake, and nonalcohol calorie intake, the ORs were 1.1 (95% CI 0.9-1.4) for fewer than six drinks per day, 1.0 (95% CI 0.4-1.4) for six to fewer than eight drinks per day, and 1.3 (95% CI 0.9-1.9) for eight or more drinks per day, and the trend in risk was not significant. No interaction was observed between alcohol drinking and sex, family history, and smoking, but the association with alcohol drinking was appreciably stronger in the elderly and in less-educated individuals. Thus this large data set was able to exclude a strong and consistent association between alcohol (mainly wine) drinking and stomach cancer risk. A nonsignificant association was observed in those who drank very heavily, but the absence of a dose-risk relationship suggests that even such a moderate association may reflect inadequate allowance for covariates or the presence of other risk factors (possibly related to diet and social class) among the heaviest drinkers.