Diet and gastric cancer mortality in Portugal was studied using a multivariate ecological model. The factors investigated over 18 districts were the relationship between gastric cancer mortality (1994–96), dietary habits, and socio-economic factors (1980–81). Mortality geographical pattern was established using age-standardized mortality rates, per capita dietary consumption of foodstuffs and nutrients was obtained from the National Alimentary Survey (1980–81), and data on socio-economic factors from the 1981 National Census. Pearson correlation coefficients and simple and multiple linear regression models were used. The mortality geographical pattern resembled a north-south gradient, and dietary habits and socio-economic factors had great variability throughout the country. The highest negative correlation coefficients between dietary consumption and gastric cancer mortality were obtained for vegetables, fruit, vitamin A and carotene consumption, and the highest positive coefficients were for rice, wine and carbohydrate consumption. No significant correlations were obtained for socio-economic factors. In multiple regression analysis, vegetable and rice consumption could account for 79% of the gastric cancer mortality variability for males, and vegetable and meat consumption could account for 69% of this variability for females. Interestingly, meat consumption was found to be protective. A mean increase of 100 g/person/day in vegetable consumption would imply a mean predicted decrease of 10 (95% CI 6–14) and 5 (95% CI 3–7) gastric cancer deaths per 100 000 persons/year, for males and females respectively, in simple regression analysis. Such a decrease represents about one-third of the mean national gastric cancer mortality rate. Therefore, an increase in vegetable consumption is strongly recommended.