The diagnostic specificity and predictive utility of the classical prognostic indicators in schizophrenia were examined in psychotic patients enrolled in the Suffolk County Mental Health Project. First-admission psychotic patients with schizophrenia (N = 96), major depression (N = 42), and bipolar disorder (N = 64) drawn from 10 facilities in Suffolk County, New York, were assessed during their initial hospitalization and at 6-month follow-up. Longitudinal consensus diagnoses were determined after the 6-month interview. The diagnostic groups shared similar background characteristics, but schizophrenics had poorer pre-morbid adjustment, longer periods of psychosis before hospitalization and more negative symptoms initially. Except for rehospitalization, schizophrenics had the worst and bipolars the best functioning at follow-up. Among the classical prognostic indicators, the best predictor of 6-month outcome for each diagnostic group was premorbid functioning.