The authors employed empirical methods to study the causes of discrepancies between clinicians' and epidemiologists' diagnoses of "cases" from the general population within the homogeneous DSM-III/DIS system. Four interviewers conducted 139 interviews using the DIS, while psychiatrists completed a DSM-III checklist, after which they could then ask any questions they wanted. All kappas exceeded 0.58. Meetings were subsequently organized with all participating psychiatrists in order to point out reasons for discrepancies between DIS diagnoses and clinical judgement. The authors came to the following conclusions: (1) DSM-III ambiguities led to discrepancies, especially when reference periods were not specified. (2) Discrepancies arose when cases were difficult: symptoms pertaining to different diagnoses or multiple diagnoses and the fact that clinicians could use volunteered comments, while interviewers were obliged to keep strictly to the schedule, contributed to discrepancies. (3) Other approaches, such as CIDI for anxiety and DISSA for depression, could improve DIS performances.