The aim of this study was to determine, in a population of medical inpatients, the sensitivity, specificity, positive and negative predictive values of two self-rating depression scales, and of the physicians' judgment, compared to a structured interview derived from the Composite Interview Diagnostic Interview (CIDI) designed to assess the diagnosis of major depressive episodes. The setting of the study was a general internal medicine inpatient ward of a French university hospital. Patients between 15 and 75 were asked to fill, within the first week of their hospitalization, two self-rating depression scales: the Beck Depression Inventory - short form (BDI-short form), a 13-item scale with established reliability and validity; and the Center for Epidemiologic Studies-Depression Scale (CES-D), a 20-item scale designed for epidemiological use in the general population, recently translated in French. Patients were then interviewed by a psychologist, blind to the results of the self-rating scales, using a slightly simplified version of the depression section of the CIDI. The ward physicians' recognition of depression was assessed six months later by a careful chart review, conducted by an investigator blind to the results of depression scales and structured interview. One hundred consecutive patients were studied: 63% were women, mean age was 53.6 +/- 16.5 years. According to the results of the CIDI, the prevalence of current major depression was 29.7%, and the lifetime prevalence of major depression was 59.8%. The correlation between BDI-short form and CES-D was +0.81. BDI-short form proved to have a slightly better acceptability than the CES-D.(ABSTRACT TRUNCATED AT 250 WORDS)