Epidemiological studies of the association between coffee consumption and dementia have yielded inconsistent results. Therefore, we investigated the association between coffee consumption and incident risk of dementia in an elderly Japanese population. 23,091 subjects aged ≥65 y living in Ohsaki City, northeastern Japan, responded to the baseline survey in 2006. Of these, we analyzed 13,137 subjects who gave informed consent and were not disabled at baseline. The outcome was the incidence of disabling dementia defined by usage of the Long-term Care Insurance database. We used the Cox proportional hazards regression model for multivariate analysis. During 5.7 y of follow-up period, we identified 1,107 cases of incident dementia. Overall, coffee consumption was significantly associated with a lower risk of incident dementia. The multivariate-adjusted HRs for the incidence of dementia according to coffee consumption categories (never, occasionally, 1-2 cups/d, and ≥3 cups/d) were 1.00, 0.73 (95% CI, 0.62-0.86), 0.72 (95% CI, 0.61-0.84), and 0.82 (95% CI, 0.65-1.02; p for trend = 0.009), respectively. In addition, this significant inverse association was more remarkable among women, non-smokers, and non-drinkers. Coffee consumption is significantly associated with a lower risk of incident dementia.