BACKGROUND: Evidence of an association between sleep duration and depression was inconsistent. METHODS: Adults who participated in the National Health and Nutrition Examination Survey (NHANES) from 2009 to 2016 were included. Multivariable logistic regression was used to evaluate the association between sleep duration and depression. RESULTS: Among the 25,962 participants (mean age 48.1 years; 49.2% male) in this study, 23,636 had a depression score <10 and 2,326 had a depression score ≥ 10. After adjustment for gender, age, race/ethnicity, education, marital status, and annual family income, BMI, alcohol status, and smoking status, hypertension, diabetes, dyslipidemia, work activity, and physical activity risk factors, participants who had short sleep duration had odds ratios (OR) of 1.86 (95% confidence interval 1.59, 2.17) and participants who had long sleep duration had OR of 1.49 (95% confidence interval 1.22, 1.83) for incident depression. Further analysis revealed a U-shaped association between sleep duration and incident depression. When sleep duration < 8 hours, increased sleep duration is associated with a significantly lower risk of incident depression (OR = 0.68 [95% CI 0.64, 0.71], P < 0.001). When sleep duration ≥ 8 hours, the risk of depression increased significantly with an increase in sleep duration (OR = 1.32 [95%CI 1.23, 1.41], P < 0.001). CONCLUSIONS: Sleep duration were independently associated with a higher incident depression. Not only insufficient sleep but excessive sleep also increase the risk of depression.