Most treatment of bipolar disorders addresses maintenance objectives. In the past 4 years, several large, blinded, randomized, placebo-controlled maintenance studies involving more than 3000 patients with bipolar I illness have been published, with analyses of background, symptomatic, and acute treatment factors contributing to maintenance effectiveness. This article summarizes these findings. Generally, indices of greater severity predict lower response rates to most monotherapy treatments. Some findings have been unexpected. Mixed mania did not predict different maintenance response to divalproex or lithium but predicted more side effects with either drug, and reduced efficacy of olanzapine or lamotrigine.