The Hamilton Depression Scale (HAMD) is the most commonly used scale for depression severity assessment and for antidepressant treatment evaluation. Alternative scales have been proposed by Bech and Rafaelsen (BRMS) and by Montgomery and Asberg (MADRS) to try to overcome the shortcomings of HAMD: they are based on different concepts of severity and different scaling procedures. Comparisons with respect to reliability, validity and ability to detect change have been performed using these scales in different samples. The BRMS proved superior. This result makes it necessary to question the usual procedure of testing the efficacy of antidepressants by means of HAMD alone. Problems in defining the severity of depression and in testing the validity of severity scales are discussed.