Patients in an effective disorders out-patient clinic were studied with four depression rating scales: the Hamilton rating scale (HRS) the Carroll rating scale (CRS) a clinical global rating of depression (CGRD) and the visual analogue scale (VAS). The overall correlations between the self ratings (CRS, VAS) and the observer ratings (HRS, CGRD) were highly significant. Both the HRS and the CRS distinguished mild from moderate, and moderate from severe depression. CRS scores increased more rapidly than HRS scores with increasing severity of depression. The concordance of self ratings and observer ratings was highest for the two structured instruments (HRS and CRS), and was lowest for the two global scales (CGRD and VAS). The global scales have the advantages of speed and simplicity, but at the cost of some reliability. Patients with non-endogenous depression had significantly increased self rating scores in comparison to patients with unipolar or bipolar endogenous depression. The correlations between the self ratings and the observer ratings were notably lower in patients with non-endogenous depression than in patients with endogenous depression. Euthymic bipolar patients rated themselves on the VAS as significantly less well than euthymic unipolar patients. The clinical and research implications of these findings are discussed.