In this prediction study 50 patients with DSM-III major depressive episode were assessed before treatment with respect to symptomatology (Newcastle and Hamilton features), diagnostic sub-types (according to the Newcastle scale and the DSM-III) and neuroendocrinology (DST and TRH-test). Subjects were given 150 mg of amitriptyline daily and the treatment response was determined on day 11 and 21. The duration of hospitalization served as a further criterion of outcome. The data suggest that neither diagnostic sub-typing nor neuroendocrinological variables were significantly related to outcome. Prediction, however, became possible using the following clinical parameters: DSM-III depressive psychotic features, DSM-III personality disorder and sudden onset of illness were significantly associated with poor treatment response and explained 34% of the outcome variance.