The clinical value of different classification systems of HIV-infection has been examined in a cohort of 99 HIV-infected individuals. AIDS-related complex (ARC), Walter-Reed (WR) stage 5 and Centers for Disease Control (CDC) group IV without AIDS cases were equally capable of identifying patients with bad prognosis: Out of 12 patients progressing to AIDS, 9 WR 5, 7 CDC IV and 7 ARC. Prognostic parameters (Erythrocyte sedimentation rate, Haemoglobin, Leucocytes, CD4-lymphocytes and Beta 2-Microglobulin) did not differ in the three groups. However, there were great differences in the frequency of patients infected by homosexual contacts and of intravenous drug-abusers. Drug-abusers were more often seen in the ARC-group, whereas there was a slight dominance of homosexuals in the WR 5-group (and a marked preponderance in the AIDS-group). Using 3 different ARC-definitions we found an ARC-frequency of 10%, 24% and 41% in our cohort. Thus the great variability of the ARC-concept could be demonstrated. In conclusion we believe that for the purpose of clinical documentation and studies the ARC-concept is unnecessary. Further studies have to evaluate other classification systems (Walter-Reed, CDC) or to establish new ones.