OBJECTIVE: Although physical activity (PA) in late life is considered a preventive factor for dementia, effects of different types of PAs on the development of dementia in early old age are unclear. This study aimed to determine the effect of leisure-time and non-leisure-time PAs on dementia risk in middle-aged and older adults during an 8-year follow-up. DESIGN: Prospective cohort study. SETTING AND PARTICIPANTS: Participants were 13,773 community-dwelling individuals aged 40-74 years who completed the baseline self-administered questionnaire survey of the Murakami cohort study in 2011-2013. METHODS: Main predictors were leisure-time and non-leisure-time (commute, occupational work, and housework) PAs as assessed by MET score (MET-hour/d). The outcome was newly developed dementia determined using a long-term care insurance database. Covariates included demographics, lifestyle, body size, disease history, and PA level. Hazard ratios (HRs) were calculated using Cox proportional hazards models. RESULTS: Mean age of participants was 59.0 (SD 9.3) years. Higher levels of leisure-time PA were associated with lower HRs (adjusted P for trend <.001), with all tertiles having significantly lower HRs (low: 0.71, 95% CI 0.51-0.99; medium: 0.59, 95% CI 0.43-0.81; high: 0.55, 95% CI 0.41-0.75) relative to the reference (zero). Higher quartiles of non-leisure-time PA were associated with lower adjusted HRs for dementia (adjusted P for trend < .001), with the second-fourth quartiles having significantly lower HRs (second: 0.73, 95% CI 0.54-0.98; third: 0.59, 95% CI 0.43-0.81; fourth: 0.55, 95% CI 0.41-0.75) relative to the lowest quartile. These associations were robust regardless of sex and age group. CONCLUSIONS AND IMPLICATIONS: Both leisure-time and non-leisure-time PAs are independently and robustly associated with a reduced risk of dementia.