The objective of this study was to examine the relationship between sleep duration and ischemic and hemorrhagic stroke in a community-based cohort. The current analysis included 95,023 Chinese participants who were free of stroke at the baseline survey (2006-2007). Cox proportional hazards models were used to calculate hazard ratios (HRs) and their confidence intervals (CIs) for stroke, according to sleep duration. After a mean follow-up period of 7.9 years, 3,135 participants developed stroke (2,504 ischemic stroke and 631 hemorrhagic stroke). The full adjusted hazard ratio (95% CI) of total stroke (with 6-8 hours of night sleep being considered for the reference group) for individuals reporting greater than 8 hours was 1.29 (1.01-1.64). More significant association between long sleep duration and total stroke was found in the elderly (HR, 1.47; 95% CI, 1.05-2.07). Compared with participants getting 6-8 hours of sleep, only women who reported sleeping more than 8 hours per night were associated with hemorrhagic stroke (HR, 3.58; 95% CI, 1.28-10.06). This study suggested that long sleep duration might be a potential predictor/ marker for total stroke, especially in the elderly. And long sleep duration increased the risk of hemorrhagic stroke only in women.