Among 4775 adults of 65 years or older and free of known cerebrovascular disease at baseline in 1989–1990, usual dietary intake was assessed using a food frequency questionnaire. In a subset, consumption of tuna or other broiled or baked fish, but not fried fish or fish sandwiches (fish burgers), correlated with plasma phospholipid long-chain n-3 fatty acid levels. Incident strokes were prospectively ascertained. During 12 years of follow-up, participants experienced 626 incident strokes, including 529 ischaemic strokes. In multivariate analyses, tuna/other fish consumption was inversely associated with total stroke and ischaemic stroke, with 27% lower risk of ischaemic stroke with an intake of once to four times per week, and 30% lower risk with intake of five or more times per week, compared with an intake of less than once per month. In contrast, fried fish/fish sandwich consumption was positively associated with total stroke and ischaemic stroke, with a 44% higher risk of ischaemic stroke with consumption of more than once per week, compared with consumption of less than once per month. Fish consumption was not associated with haemorrhagic stroke.