This study examines the use of the Center for Epidemiological Studies-Depression Scale (CES-D) in a sample of elderly, medically ill inpatients. Seventy-six individuals completed the CES-D and a psychiatric interview, from which DSM-III-R diagnoses of depression were obtained. Analyses of sensitivity and specificity indicated that use of an alternative scoring method which more closely approximates current diagnostic criteria for depression may improve the predictive power of the test. Employment of stringent cut-scores was not supported, as sensitivity was compromised. Item analyses demonstrated that seven of the CES-D items failed to discriminate major, minor and non-depressed patients, and that several of these items tapped somatic symptoms. These findings suggest that the validity of the CES-D may be compromised when used with elderly medical patients, and modifications for its use are recommended. © 1997 by John Wiley & Sons, Ltd.