The authors compared patients meeting widely accepted criteria for the diagnosis of schizoaffective disorder, manic type, with patients meeting rigorous criteria for manic disorder and schizophrenia, using three methods of validation: family history, short-term treatment response, and long-term outcome. No significant differences were found between patients with manic disorder and schizoaffective disorder. However, consistent and often highly significant differences separated patients with schizophrenia from those with manic disorder and schizoaffective disorder. The findings suggest that schizoaffective disorder, as currently defined, is not a valid and independent entity. The authors suggest that psychotic disorders not diagnosable as manic- depressive illness or schizophrenia and without apparent organic basis would best be called "undiagnosed" or "atypical" psychosis. Further, while proposals for new diagnoses or for subtyping of schizophrenia or manic-depressive illness should be encouraged, these should undergo rigorous screening for validity before being accepted into clinical use.