Results of previous studies on fish intake and stroke risk have been inconclusive. Different stroke types have often not been separated. Our aim was to elucidate whether intake of fish, Hg or the sum of proportions of fatty acids EPA (20 : 5n-3) and DHA (22 : 6n-3) influence the risk of haemorrhagic or ischaemic stroke. Within a population-based cohort from a community intervention programme, 369 stroke cases and 738 matched controls were identified and included in the present nested case–control study. Information on fish intake had been recorded at recruitment, i.e. before diagnosis. Hg levels were determined in erythrocyte membranes, also collected at recruitment, and the relative content of fatty acids was measured in erythrocyte membranes or plasma phospholipids. The results showed that in women there was a non-significant decrease in stroke risk with increasing fish intake (OR 0·90 (95 % CI 0·73, 1·11) per meal per week). The risk in women differed significantly (P = 0·03) from that in men, in whom the OR for stroke rose with increasing fish intake (OR 1·24 (95 % CI 1·01, 1·51) per meal per week). The corresponding risk in men for Hg was 0·99 (95 % CI 0·93, 1·06), and for the sum of proportions of EPA and DHA 1·08 (95 % CI 0·92, 1·28). We conclude that the relationship between stroke risk and fish intake seems to be different in men and women. Increased levels of EPA and DHA do not decrease the risk for stroke and there is no association between stroke risk and Hg at these low levels.