97 nonpsychotic consecutive day patients were diagnosed by the axis 1 and 2 in the DSM-III and DSM-III-R system, and their treatment response during their stay was measured by the Health Sickness Rating Scale. The interrater reliability was equally good for both diagnostic systems. On axis 1, there were only minor differences between DSM-III and DSM-III-R. On axis 2, the frequency of schizotypal disorder was reduced by 40% and the frequency of histrionic disorder by two-thirds. The number of schizoid disorders increased from zero to five. Of the DSM-III schizotypals who lost this diagnosis in DSM-III-R (n = 8), 4 got a new diagnosis of schizoid personality and 4 maintained their borderline diagnoses. In DSM-III-R there was a sharper demarcation between patients with severe and nonsevere personality disorder with regard to treatment outcome, indicating an increased validity of these categories. There was also a sharper demarcation between borderline versus histrionic and schizotypal, and between schizotypal and schizoid diagnoses.