In order to assess the concordance between self-rating and clinician's assessment tools of depression, as well as factors involved in the differences between auto and hetero evaluation, 198 depressed in-patients were assessed at admission and at discharge using the Montgomery Asberg Depression Rating Scale (10-item version, MADRS) and the self-rating scale Symptoms CheckList (90-item version, SCL-90). We found that about 18% of patients overestimated and about 15% underestimated their depressive symptomatology (SCL-90 depression subscale) relative to the psychiatrist's assessment. Logistic regression analysis showed that the presence of personality disorders and previous history of psychiatric disorders predicted the overestimating group. Discriminant analysis showed that approximately 75% of participants were correctly classified when previous history of psychiatric disorders, presence of personality disorders and age were entered separately into the equation.