The usefulness of several historical, clinical and biological variables as possible predictors of outcome was tested in a sample of patients with a cross-sectional diagnosis of schizoaffective disorder, depressed type. Four historical items were found to be successful: a family history of chronic schizophrenia, the occurrence of schizophrenic symptoms at some stage of the illness in the absence of depression and an onset of the index episode as exacerbation of previous symptoms (all associated with a relatively poor outcome), and a personal history of previous manic episodes (associated with a relatively good outcome). The various aspects of the clinical picture during the index episode, as well as the response on dexamethasone suppression test, were not found to have any predictive value. These findings confirm that, in patients with a cross-sectional diagnosis of schizodepressive disorder, the previous course of the illness is of crucial importance for prognosis, and support the usefulness of a multiaxial classification of schizoaffective states, taking into account not only cross-sectional symptomatology but also course.