This study examined the inter-observer reliability and validity of the Center for Epidemiologic Studies Depression Scale (CES-D) as a measure of depressive symptomatology in stroke patients, and its utility as a screening tool for depression in this population. The CES-D Scale is a brief questionnaire originally designed for use in community surveys. Twenty-seven non-aphasic patients enrolled in the Stroke Data Bank at the University of Maryland were interviewed by a research nurse using the CES-D. On the same day, each patient was independently evaluated by a research assistant using a psychiatric battery for depression and measures of cognitive, physical, and social functioning. Forty-one percent (11/27) of the patients were depressed according to clinical criteria for major or minor depression. With a cutpoint corresponding to the upper (most severe) 20% in community surveys, the CES-D Scale picked up 73% (8/11) of the depressed patients. In this sample no nondepressed patient scored over 16 on the CES-D (no false positives). The CES-D Scale scores correlated significantly with the other depression measures (r = .57 to r = .82, p less than .002) and did not correlate with the measures of cognitive, physical, or social functioning. Based on 24 patients who received a CES-D Scale score from both the nurse and the research assistant, inter-rater reliability was high (r = .76, p less than .001). Thus, the CES-D was found to be reliable and valid as a screening tool for assessing depression in stroke patients.