Background-To date, no prospective studies have examined the association between serum homocysteine levels and the risk of stroke and stroke subtypes in Asian populations. Methods and Results-A prospective, nested, case-control study of Japanese subjects 40 to 85 years of age was conducted by using frozen serum samples from 11 846 participants in cardiovascular risk surveys collected from 1984 to 1995 for one community and 1989 to 1995 for the other two communities. By the end of 2000, we identified 150 incident strokes, the subtypes of which were confirmed by imaging studies. Three control subjects per case were selected by matching for sex, age, community, year of serum storage, and fasting status. Serum total homocysteine levels were measured by high-performance liquid chromatography. Compared with control subjects, total (nϭ150), hemorrhagic (nϭ52), and ischemic (nϭ98) strokes had higher geometric mean values of total homocysteine and higher proportions of homocysteine Ն11.0 mol/L. The multivariate odds ratios (95% CI) for highest (Ն11.0 mol/L) versus lowest quartiles (Ͻ7.0 mol/L) of homocysteine after adjustment for body mass index, smoking, alcohol intake, hypertension, serum total cholesterol, and other cardiovascular risk factors were 2.99 (1.51 to 5.93) for total stroke, 3.89 (1.60 to 9.46) for ischemic stroke, 3.36 (1.27 to 8.90) for lacunar infarction, and 1.63 (0.44 to 6.00) for hemorrhagic stroke. The respective multivariate odds ratios associated with a 5-mol/L increase in homocysteine were 1.40 (1.09 to 1.80), 1.52 (1.07 to 2.14), 1.48 (1.01 to 2.18), and 1.10 (0.76 to 1.59). The excess risk of total and ischemic strokes did not vary significantly according to sex, age, smoking status, or hypertensive status. Conclusions-High total homocysteine concentrations were associated with the increased risk of total stroke, more specifically ischemic stroke and lacunar infarction, among Japanese men and women.