In a five year follow-up study of 57 hospitalized chronic schizophrenics we investigated the development of symptoms and syndromes, taking into account different nosological concepts. We could demonstrate that after a careful methodological training Leonhard's differentiated classification of schizophrenia is highly reliable. Analyses of the long-term course of individual syndromes showed that a dichotomy into positive and negative symptoms does not provide a reliable prognosis. Within the Leonhard classification, however, a remarkable homogeneity was evident with regard to course and outcome of various schizophrenic subtypes. Investigating the biography of the kinships we found further indication of nosologic heterogeneity, particularly between unsystematic (high hereditary loading) and systematic schizophrenias (almost no hereditary loading). However, the strategy of a simple dichotomy into familial/sporadic cases appears to have too many methodological shortcomings. Further, the study showed that, in this cohort, neuroleptic long-term treatment did not decisively influence the development of residual states, with or without residual marked positive symptoms. It is suggested that the Leonhard classification is a promising concept for further psychopathological as well as biological investigations in psychiatry.