OBJECTIVES: To define criteria for symptoms remission in depression for reduced versions of the HDRS. METHODS: The discriminative validity of the HDRS (HDRS-21 and 17 items) and its shorter scales [Bech Melancholia Scale (BMS-6); Maier & Philips Severity Scale (MPSS-6); Gibbons et al Depression Scale (GDS-8); Evans et al Depression Scale (EDS-6)], was assessed against the Clinical Global Impression of severity (CGI) in a Spanish multicenter study. The study included 168 depressive patients in ambulatory care. Of these, 118 patients were considered as clinically unstable (either because of presenting a new/recurrent disease episode or because of needing an adjustment/change of treatment). After six weeks, those patients were reassessed by the HDRS and the CGI. The best cut-off points to discriminate the criterion of clinical remission (CGI score = 1) were found by using Receiver Operating Characteristic analyses (ROC). The accuracy of the different versions was assessed by analysing the area under the ROC curves (AUC). RESULTS: All versions discriminated at baseline the severity of depression according to the CGI criterion (all p-values <0.005 by oneway ANOVA corrected for multiple comparisons). Also, all versions discriminated the severity of depression (remission versus no remission) at six weeks of follow-up (all p-values <0.0001 by t-test analyses). The ROC analyses suggested the following cut-off points to identify remission criteria in our sample (HDRS-21: