The factorial and discriminative validity of the Hopkins Symptom Check List SCL-90-R were examined in the light of criticism that clinical self-rating scales primarily express a general distress factor. In a population of 899 psychosomatic patients, high intercorrelations were found between the individual dimensions of the SCL-90-R. A subsequent Principal Components Analysis obtained 9 factors which were markedly less interdependent than those in the original version. The ability of the questionnaire to distinguish between patients with dysthymia, anxiety disorders and anorexia nervosa was examined. The average hit rate in the discriminant analysis was 67% using the original version and 74% with the proposed new factorial structure of the SCL-90-R, confirming the discriminative validity of the inventory. The present results as well as earlier studies suggest that the factor ‘anxiety’ should be included in the factor ‘phobic anxiety’, the factors ‘paranoid ideation’ and ‘psychoticism’ should be reformulated, and a new factor ‘sleep disturbances’ should be added to the original version of the SCL-90-R.