The relation between initial clinical symptomatology and long-term course of illness appears to be reliable. Prognostic models, as well as an eyperienced physician, can predict long-term risk of schizophrenic defects from a few items.The prognostic models also seem to be valid for international comparisons of prognosis. The improved prognosis of schizophrenia over the last 30 years is ascertained. The series of about 4,500 cases provides a sufficiently large number of patients for the analyses of well-defined subgroups. This has been done especially for schizophrenia, using the Leonhard classification.