The authors compared DSM-III and DSM-III-R definitions of schizophrenia among 532 inpatients treated in a long-term residential setting and reevaluated an average of 15 years later. Largely by excluding those with nonbizarre delusions (somatic, grandiose, or religious) without hallucinations, DSM-III-R reduced the number of patients diagnosed with schizophrenia by 10%. With the exception of the sign and symptom variables used to define them, the DSM-III schizophrenic patients included (N = 164) and excluded (N = 18) by DSM-III-R did not differ with respect to demographic, premorbid, or long-term outcome characteristics. The authors argue that frequent changes in diagnostic schemes in the absence of evidence of improved validity are likely to impede progress in research.