Moderate hyperhomocysteinaemia is common in the general population and has been linked with cardiovascular disease. However, there are no data from prospective, population-based studies. We examined the association between serum total homocysteine (tHcy) concentration and stroke in a nested case-control study within the British Regional Heart Study cohort. Between 1978 and 1980 serum was saved from 5661 men, aged 40-59 years, randomly selected from the population of one general practice in each of 18 towns in the UK. During follow-up to December, 1991, there were 141 incident cases of stroke among men with no history of stroke at screening. Serum tHcy was measured in 107 cases and 118 control men (matched for age-group and town, without a history of stroke at screening, who did not develop a stroke or myocardial infarction during follow-up). tHcy concentrations were significantly higher in cases than controls (geometric mean 13.7 [95% CI 12.7-14.8] vs 11.9 [11.3-12.6] mumol/L; p = 0.004). There was a graded increase in the relative risk of stroke in the second, third, and fourth quarters of the tHcy distribution (odds ratios 1.3, 1.9, 2.8; trend p = 0.005) relative to the first. Adjustment for age-group, town, social class, body-mass index, hypertensive status, cigarette smoking, forced expiratory volume, packed-cell volume, alcohol intake, diabetes, high-density-lipoprotein cholesterol, and serum creatinine did not attenuate the association. These findings suggest that tHcy is a strong and independent risk factor for stroke.