In a German multicenter treatment study, 354 patients with schizophrenia and schizoaffective disorder were followed for 2 years. The data collected were taken as a basis for the present predictor study. For the first time, the technique of classification and regression tree (CART) analysis has been employed for this purpose. CART yielded informative data and appeared to be a useful instrument in predictor research. On the outcome variables "relapse" and "rehospitalization," significant predictor variables were found in several areas: neuroleptic treatment, onset and previous course (precipitating factors, first manifestation, hospitalization in the preceding year, suicide attempts), psychopathology (residual type, schizoaffective disorder), social adjustment (marital status, employment, intensity of life, Phillips score), previous life experiences (traumatic experiences and psychiatric or developmental disturbances in childhood), and biology (gender, age). Our investigation confirmed the generally prevalent views regarding the value of neuroleptic treatment, the multifactorial etiology, and the vulnerability stress model of schizophrenia.