Abstract: Objectives: The Centre for Epidemiologic Studies Depression Scale (CES-D) is a self-rating scale for the measuring depressive symptoms in community settings. It is commonly used as a screening instrument for possible depression. Factor analytic studies have shown that the 20 items of CES-D represent depressed affect, positive affective (reversed items), somatic complaints, and interpersonal concerns (Radloff, 1977). Although it is one of the most commonly used questionnaires in the world, limited studies have focused their attentions on the factor structures of the CES-D among the groups of Chinese elderly adults. The objective of the present study was to examine the factor structures of CES-D in Chinese elderly populations. Method: In a national sample of 4903 elderly adults recruited from major Chinese mainland cities, the CES-D was assessed. Results: a) Confirmatory factor analysis for 20 items of CES-D demonstrated that the four- factor structure identified by Radloff (1977), as well as two three- factor structures suggested by Guarnaccia et al. (1989) and Yen et al. (2000) fit the present data well with RMSEA, SRMR＜0.08，NNFI、GFI and AGFI＞0.90, as well as CFI＞0.95, which indicated that these factor structures could be replicated in Chinese elderly population; b) The correlations between positive affect factor which consist of four reverse scoring items and the other factors were too low to support a common higher order factor, the range of rs was from 0.35 to 0.46, which was smaller than the cut-off value of .50; c) Based on point b, under the conditions that the total score is used to indicate individuals’ levels of depressive symptoms, the four items of positive affect should not be included in the CES-D. In the following analysis, we tried to test the factor structures of CES-D with exclusion of positive factor and found positive affect factor only made limited contributions to the model fits. Combining both original definition and fit indices, Radloff’s factor structure was chosen as the most suitable for the CES-D consist of remaining 16 items RMSEA=.067, SRMR=.033, CFI=.98, NNFI=.98, and GFI=.94; d) Measurement invariance was established across gender for the factor structure of the remaining 16 items. Conclusion: a) Among Chinese elderly adults, the CES–D may provide a better measure of depressive symptoms as a 16-item scale with the positive affect factor left out ; and b) The factor structure identified by Radloff fit the 16-item scale best.