OBJECTIVE: There is accumulating evidence that involvement in leisure activities may be related to risk of dementia; however, there is no consensus concerning the underlying mechanism of this association. Hypothesizing that leisure activities may contribute to cognitive reserve (CR), we examined the association between leisure activities and risk of incident dementia and its subtypes within a general population sample, categorizing leisure activity as stimulating, passive, physical, and social. The possibility that these associations may be driven by other proxies of CR was also examined. METHODS: Analyses were carried out on 5,698 dementia-free participants aged 65 and over included in the Three-City cohort study in Dijon and Montpellier (France) in 1999-2001. Hazard ratios (HR) were calculated for incident dementia and its subtypes (mixed/vascular dementia and Alzheimer disease) in relation to category of leisure activity. RESULTS: Stimulating leisure activities were found to be significantly associated with a reduced risk of dementia (n = 161, HR = 0.49, 95% confidence interval [CI]: 0.31; 0.79) and Alzheimer disease (n = 105, HR = 0.39, 95% CI: 0.21; 0.71) over the 4-year follow-up 1) independently of other proxies of CR, 2) after adjusting for vascular risk factors, depressive symptoms, and physical functioning, and 3) independently of other leisure activities. Furthermore, no significant association was found with other leisure activities and dementia after controlling for the potential confounders. CONCLUSION: Our findings support the hypothesis that cognitively stimulating leisure activities may delay the onset of dementia in community-dwelling elders.