Background: Treatment resistant depression is a common clinical problem. Studies have shown that a large number of patients with depression do not have a satisfactory clinical outcome in spite of adequate trials of antidepressant drugs. In this study, we investigated demographic and clinical characteristics, diagnostic subtypes, and illness outcome of patients with resistant depression and a history of escape of response to adequate trials of at least two antidepressants for a previous episode. Method: Sixty-one patients who were seen consecutively at a mood disorders clinic with the diagnosis of “unipolar” treatment resistant depression, and followed up for at least one year, were interviewed using the Structured Clinical Interview for DSM-IV. Prospectively collected data including the occurrence of episodes of hypomania, and supplemental information from family members on illness course were also used for purposes of diagnostic re-evaluation. Results: At intake, 35% of the patients were diagnosed as having a bipolar disorder. At follow-up, there was a 59% prevalence of bipolar disorder. Of the patients with major depressive disorder, 52% were subsequently classified as having bipolar spectrum disorder. The most important finding was that 80% of patients were found to show evidence of bipolarity. Moreover, the most common change in medication was a switch to mood stabilizers. CGI ratings showed significant improvement in functioning from the time of initial consultation. Limitations: This was a naturalistic study, and the data were collected in a non-blind fashion. Conclusions: The findings suggest that the majority of cases of unipolar treatment resistant depression, occurring in the context of loss of antidepressant response, have a bipolar diathesis.