An overview of the literature and some findings of the author's own empirical studies lead to the general conclusion that observer rating should be chosen as the principle outcome criterion in antidepressant drug trials. In various situations, however, self-rating scales can provide additional important information for therapy evaluation. The cross-sectional confirmity of observer ratings and self ratings of depression varies to a greater or lesser extent, depending on various background variables. With respect to longitudinal observation the relationship is much closer. Sensitivity for change can be attested not only to observer ratings but also to self ratings.