Of the placebo-controlled maintenance studies conducted in bipolar disorder, few have enrolled patients who present depressed. In fact, only lithium and lamotrigine have been studied over the long term with placebo-controlled designs in recently manic and recently depressed bipolar patients. Given the magnitude of the unmet medical need and the data suggesting that symptomatic patients with bipolar disorder spend the majority of their time depressed, this is unfortunate. Our review of the pre-lithium literature and more recent publications suggests that mood state at study entry predicts the polarity of relapse and the response to treatment. Accordingly, a need exists to enroll recently depressed patients in maintenance studies to elucidate the complete spectrum of efficacy of putative mood stabilizers and improve the long-term treatment of bipolar depression. Patients presenting depressed for a maintenance study tend to relapse into depression; those presenting manic, into hypomania/mania/mixed states. This is particularly true during the first several months of the randomized treatment. The polarity of the index episode tends to predict the polarity of relapse into a subsequent episode in a ratio of about 2:1 to 3:1. We conclude that putative mood stabilizers must be tested in recently manic and recently depressed patients to determine their spectrum of prophylactic efficacy.