OBJECTIVE It is currently assumed that there are no important differences between the clinical presentations of unipolar and bipolar depression. Failure to distinguish bipolar from unipolar depression may lead to inappropriate treatment and poorer outcomes. We hereby compare unipolar and bipolar depressed subjects, in order to identify distinctive clinical specificities of bipolar depression. METHODS Two independent samples of depressed patients (unipolar and bipolar) were recruited, with 55 patients in one sample, and 49 in the other. In both samples, unipolar and bipolar patients were compared on a broad range of parameters, including sociodemographic characteristics, comorbidities, Montgomery and Asberg Depression Scale (MADRS; assessing depression severity), CORE (assessing psychomotor disturbance) and Bipolar Depression Rating Scale (assessing specific bipolar depression symptoms). RESULTS Results were similar in both samples. MADRS scores were similar in bipolar and unipolar subjects (median score 33 vs 34; p=0.74). On the CORE, there was a trend to higher scores among the bipolar subjects. BDRS scores were higher in bipolar than in unipolar subjects (median score 33 vs 27; p<0.001). The difference was particularly marked on the "mixed" subscale of the BDRS. We tested the ability of the mixed subscale of the BDRS to distinguish bipolar from unipolar depression, using different cut off points: a cut off point of 3 can predict bipolar depression, with a sensibility of 62% and a specificity of 82%. CONCLUSIONS Presence of mixed symptoms during a depressive episode is in favour of bipolar depression. The BDRS scale should be integrated in a probabilistic approach to distinguish bipolar from unipolar depression.