Studies on the intake of different types of carbohydrates and long-term mortality are sparse. We examined the association of starch, total and each type of sugar and free sugars with the risk of total and cause-specific mortality in a cohort of the general population in Japan. Study subjects were 29 079 residents from the Takayama Study, Japan, who responded to a self-administered questionnaire in 1992. Diet was assessed by a validated FFQ at the baseline. Mortality was ascertained during 16 years of follow-up. We noted 2901 deaths (974 cancer related and 775 cardiovascular related) in men and 2438 death (646 cancer related and 903 cardiovascular related) in women. In men, intake of starch was inversely associated with total mortality after controlling for covariates (hazard ratio (HR) for the highest quartile v. lowest quartile: 0·71; 95 % CI 0·60, 0·84; Ptrend < 0·001). Intakes of total sugars, glucose, fructose, sucrose, maltose and free and naturally occurring sugars were significantly positively associated with total mortality in men (HR for the highest v. lowest quartile of total sugar: 1·27; 95 % CI 1·12, 1·45; Ptrend < 0·0001). Similar relations were observed for cardiovascular mortality and non-cancer, non-cardiovascular mortality in men. In women, there was no significant association between any type of carbohydrates and mortality except that intake of free sugars was significantly positively associated with total and non-cancer, non-cardiovascular mortality. Data suggest that the high intake of starch reduces mortality, whereas the high intake of sugars, including glucose, fructose and sucrose, increases mortality in Japanese men.