Objective To examine the association of fish intake with the subsequent risk of impaired glucose tolerance and diabetes mellitus (glucose intolerance). Research Design and Methods In 1971, information about food intake was obtained by the cross-check dietary history method on 175 men and women aged 64-87 yr who were normoglycemic and free of clinical diabetes. During the follow-up period from 1972 to 1975, an oral glucose tolerance test was performed annually, and in 59 of 175 elderly people a diagnosis of glucose intolerance was made at least once. Results In 1971, ~ 60% of the subjects usually ate fish, with a mean daily intake of 24.2 g. In fish eaters, the incidence of glucose intolerance was significantly lower compared with nonfish eaters (odds ratio [OR] 0.40, 95% confidence interval [CI] 0.21-0.77). With logistic regression analysis, this inverse association could not be explained by taking into account age and sex or possible confounding baseline characteristics, such as the prevalence of myocardial infarction, body mass index, energy intake per kilogram body weight, or intake of carbohydrates (OR 0.47, 95% CI 0.23-0.93). Baseline characteristics of the oral glucose tolerance test and serum triglyceride levels could also not account for this result. Conclusions These results suggest that, in an elderly population, the habitual consumption of a small amount of fish may protect against the development of impaired glucose tolerance and diabetes mellitus.