Objective: No study has examined the diagnostic validity of the Center for Epidemiologic Studies Depression Scale (CES-D) in the Chinese elderly. This study aims to determine appropriate cutoffs for the 20- (CESD-20) as well as a ten-item (CESD-10) version of the instrument. Data were also provided, based on simulated scoring, for the diagnostic performance of the scales when using dichotomous instead of 4-point rating scales. Methods: Three hundred and ninety eight persons aged 60 +referred for psychiatric asssessment by a physician were administered the CES-D as well as given an independent psychiatric assessment. A spectrum of depression diagnosis as the criterion was used to assess the diagnostic validity of the CES-D. Results: The ten and the 20-item version of the CES-D, regardless of scoring method, produced essentially identical performance indices. The optimal thresholds were 12 and 22 for CESD-10 and CESD-20 respectively, and based on these thresholds, sensitivity, specificity, positive predictive value and negative predictive value were 0.76, 0.55, 0.57 and 0.74 for CESD-10, and 0.75, 0.51, 0.55 and 0.72 for CESD-20. With both ends of the rating scale collapsed to create dichotomous items, the optimal thresholds became 4 for CESD-10 and 7 for CESD-20, and the correponding performance indices were 0.67, 0.58, 0.56 and 0.69 for CESD-10, and 0.70, 0.58, 0.57 and 0.70 for CESD-20. Conclusions: The ten-item version can be used in lieu of the 20-item version, and a dichotomous response format would probably work as well as the original four-point format, in order to simplify administration for elderly persons.