Whole soy foods, as well as their components, including protein and isoflavones, have garnered attention because they may have beneficial effects against diabetes. We examined associations between the intake of soy foods, soy protein, and soy isoflavones and the risk of diabetes in the Japanese population. This prospective cohort study included 13,521 residents (5883 men and 7638 women; 35-69 y old) of Takayama City, Japan. The subjects responded to a self-administered baseline questionnaire in 1992 and to a follow-up questionnaire seeking information about diabetes in 2002. Their mean ± SD body mass index was 22.6 ± 2.6 kg/m2 (men) and 22.1 ± 2.7 (women). The intakes of total soy foods, fried soy foods, nonfried soy foods, soy protein, and soy isoflavones were estimated through the use of a validated food-frequency questionnaire administered in 1992. Associations between soy intake and the risk of diabetes were evaluated through the use of Cox proportional hazards models incorporating age, education level, physical activity, smoking status, alcohol consumption, history of hypertension, use of vitamin supplements, menopausal status, and dietary factors including glycemic load, total energy, total fat, meat, fruit, vegetables, and coffee. During a 10-y follow-up, 438 participants reported physician-diagnosed diabetes. Women in the highest tertile of intakes of total soy foods, fried soy foods, nonfried soy foods, soy protein, and soy isoflavone had significantly lower HRs, after controlling for covariates, than those with the lowest intakes. For example, HRs were 0.45 (95% CI: 0.30, 0.68; P-trend <0.001) for total soy food intake. In men, there were no significant associations between soy intake and the risk of diabetes. These results suggest that a high soy intake may be associated with a lower risk of diabetes in Japanese women.