A definite (anchored) and a semidefinite (semi-anchored) questionnaire version of the Hamilton Depression Rating Scale (HDS) and the Bech-Rafaelsen Melancholia Scale (MES) were compared with the HDS/MES by observer-rating and self-rating of 24 patients fulfilling the DSM-3R criteria for major depressive disorder. Both types of questionnaire showed substantial agreement with the observer scale from which they were derived. The sum scores were for the definite questionnaires and the corresponding observer scales closely similar whereas the sum scores of the semidefinite questionnaires were significantly higher than the sum scores of the corresponding observer scales. These results indicate that patients' 'halo' effect may be avoided by using definite scaling criteria for self-rating. Thus, of the two versions of questionnaires the definite versions are recommended.