The aim of this study was to evaluate the diagnostic validity of the Korean version of the Geriatric Depression Scale (GDS-K) and its short form among elderly psychiatric patients. After three preliminary trials, the authors translated the GDS, including the Geriatric Depression Scale—Short Form (SGDS) into Korean. The GDS-K, the Korean version of the SGDS (SGDS-K), Hamilton Rating Scale for Depression (HRS-D), and the Center for Epidemiologic Studies Depression Scale (CES-D) were administered to 154 elderly psychiatric patients. In addition, the Diagnostic Interview Schedule (DIS-III-R) was administered independently to diagnose DSM-III-R major depression. Reliability and validity test, optimal cutoff point estimation, and receiver operating characteristics (ROC) curve analysis were performed to investigate the diagnostic validity of the GDS-K and SGDS-K. Internal consistency–reliability and concurrent validity of the GDS-K and SGDS-K associated with other depression scales (HRS-D, CES-D) were excellent. Content validity and discriminant validity, which differentiate DSM-III-R major depression from nonmajor depression was also good. We suggest a score of 16 as the optimal cutoff point of GDS-K for screening DSM-III-R major depression among clinical populations and a score of 8 as optimal cutoff score of SGDS-K. ROC curve analysis indicated high diagnostic validity for both GDS-K and SGDS-K in assessing DSM-III-R major depression. Moreover, we found that the GDS-K and SGDS-K were highly correlated ( r=.9522). This finding suggests that the SGDS-K can be used as an adequate substitute for the GDS-K. The GDS-K and SGDS-K proved valid and reliable case-finding tools for screening DSM-III-R major depression among the elderly psychiatric patients in Korea. The relatively high cutoff points of both the GDS-K and SGDS-K require further evaluation from the viewpoint of culturally determined response style in elderly Koreans.