A development version (N I M H 67) of the Hamilton depression rating scale with 26 items was scored on 125 depressed inpatients recently hospitalized for a major non schizophrenic depressive illness. The data were obtained before any antidepressant treatment or after an adequate wash out period. A principal component factorial analysis with Varimax rotation was used. Computer simulations showed that only 2 factors were really outside the non significant range. A criterion is defined to allow the choice of an optimal saturation threshold for inclusion of an item in a factor. It assigns 25 of the 26 items of the scale to one of 2 independent factors: depression and anxiety-somatization.