The test-retest reliability of DSM-IIIR diagnoses for affective and anxiety disorders was determined under clinical routine conditions in a psychiatric outpatient department. The sample consisted of 60 patients, and the Munich Diagnostic Checklists (MDCL) were administered for diagnostic evaluation and classification. Each subject was independently examined by two of four participating diagnosticians (two psychiatrists, two psychologists). Acceptably high levels of agreement were indicated by several statistics (including kappa) for most disorders. Reliability was analyzed for diagnoses, subclassifications, and symptoms. Reduced agreement was found only for dysthymia, agoraphobia, and social phobia. Major causes were information variance and weaknesses of operationalization. Overall results were satisfactory when compared with other reliability studies.