Three problems in the prediction of the long-term outcome of schizophrenia are illustrated by the results of three separate studies: The limitations of the possibility of generalizing results, the limits being due to the given historical and sociocultural settings. This determines the results of research. The relationships between different types of predictors and targets. The disease-related specificity of predictors. A prospective 1-year follow-up study comparing 100 schizophrenic patients in a rural region with 200 schizophrenic patients in an urban region shows regional differences in outcome criteria, such as rate of hospitalization. The prognostic significance of the various predictors is also different in the two regions. A follow-up study of 70 schizophrenic patients, who were continuously treated with neuroleptic drugs in our outpatient clinic after hospital discharge for an average of 14 years, shows a relatively good outcome. Several outcome dimensions (rehospitalization rate, symptoms, social and work adjustment, self-ratings) are partly mutually independent. The various outcome dimensions are predicted by different predictor patterns. A prospective follow-up study of 86 schizophrenic patients compared with 75 patients with other psychiatric diagnoses confirms the finding of the partly low intercorrelations of the different outcome criteria. The study additionally shows that the predictors of some outcome dimensions, such as work adjustment, are non-specific in respect of diagnosis.