The long-term course or natural history of schizophrenia is correlated with differing diagnostic criteria and commonly agreed upon prognostic variables. A review of 38 long-term followup studies of hospitalized schizophrenics reveals that unspecified or Kraepelinian-type schizophrenia has a much worse prognosis than atypical schizophrenia, schizoaffective psychosis, reactive psychosis, or other good premorbid types. Diagnoses based on longitudinal as well as cross-reactional data are more predictive of outcome than cross-sectionally based diagnoses. Drug and psychosocial treatment results must be evaluated in terms of prognostic variables, many of which are incorporated in some currently employed diagnostic criteria. There is no firm evidence that maintenance medication is indicated in some good prognosis patients. The paucity of long-range followups, the inadequacies of outcome assessments, and diagnostic disagreements limit our understanding of the effects of drug treatment, a treatment which is not without dangerous neurological side effects in many patients.