Objective: To assess the relationship between dietary lipids and incidence of cerebral infarction in a Japanese rural population. Design: A cohort study from July 1977 through December 1992. Setting: Akadani-Ijimino (A-I) district, Niigata Prefecture, Japan. Subjects and methods: All the residents, 1, 182 men and 1, 469 women, aged 40 years and over. Out of these members, 954 men and 1, 329 women who were initially free of stroke completed a semiquantitative food frequency ques-tionnaire in 1977, and were then subjected to a follow-up for 15.5 years. The occurrence of stroke was determined by the annual follow-up exam-ination and registry. Dietary lipid was adjusted for total energy or fat intake by the residual method. Sex- and age-stratified and blood pressure-and atrial fibrillation-adjusted relative risk for cerebral infarction was estimated by the Cox proportional hazard model. Results: There were 75 new cases of cerebral infarction during the observation period. The relative risk for cerebral infarction was less than one in the highest quartile level of total fat, saturated fatty acids (S), Keys score and west-ernized dietary pattern: 0.68-0.94. It ranged between 1.36 and 1.57 in the highest level of polyunsaturated (P), n-3 and n-6 fatty acids, and P/S ratio. Conclusion: This study suggests the possibility that the traditional Japanese diet, very low fat intake, was likely to increase the risk of stroke through the low level of serum cholesterol as an intermediary factor.