OBJECTIVE: The present study investigated the concordance of clinical and research-based DSM-III-R diagnoses in community, public, and university hospital first-admission patients. In addition to demographic characteristics, information and criterion variance were assessed as explanations for the diagnostic disagreements. METHOD: As part of the Suffolk County (New York) Mental Health Project, 223 first- admission subjects with a psychotic disorder were interviewed with the Structured Clinical Interview for DSM-III-R, and consensus diagnosis was made by two project psychiatrists. Clinical diagnoses were abstracted from discharge summaries obtained subsequent to the research diagnoses, and reasons for disagreement between the two diagnoses were determined. RESULTS: Moderate overall agreement between facility and research diagnoses was found, with highest agreement with the university facility, lowest with the public facilities, and intermediate agreement with the community facilities. Demographic variables were not significantly associated with diagnostic discordance. Apparent reasons for diagnostic disagreement included evidence of variability in information available to clinicians and research psychiatrists (N = 39 or 48% of cases with disagreement), as well as in clinical judgment in the application of DSM-III-R criteria (N = 42 or 52% of cases with disagreement). CONCLUSIONS: Considerable differences between facility and research diagnoses remain, especially in the public and community sectors; these differences can be attributed to information and criterion variance. Longitudinal follow- up is necessary to establish the predictive validity of the initial clinical and research diagnoses. Future research should also address other possible reasons for these discrepancies.