Objective: To explore an independent association between self-reported sleep duration and cause-specific mortality.Methods: Data were obtained from the Multiethnic Cohort Study conducted in Los Angeles and Hawaii.Results: Among 61,936 men and 73,749 women with no history of cancer, heart attack or stroke, 19,335 deaths occurred during an average 12.9year follow-up. Shorter (≤5h/day) and longer (≥9h/day) sleepers of both sexes (vs. 7h/day) had an increased risk of all-cause and cardiovascular disease (CVD) mortality, but not of cancer mortality. Multivariable hazard ratios for CVD mortality were 1.13 (95% CI 1.00-1.28) for ≤5h/day and 1.22 (95% CI 1.09-1.35) for ≥9h/day among men; and 1.20 (95% CI 1.05-1.36) for ≤5h/day and 1.29 (95% CI 1.13-1.47) for ≥9h/day among women. This risk pattern was not heterogeneous across specific causes of CVD death among men (Phetero 0.53) or among women (Phetero 0.72). The U-shape association for all-cause and CVD mortality was observed in all five ethnic groups included in the study and by subgroups of age, smoking status, and body mass index.Conclusion: Insufficient or excessive amounts of sleep were associated with increased risk of mortality from CVD and other diseases in a multiethnic population.