Bipolar disorder is characterized by a chronic and fluctuating course of illness. Although nonadherence to pharmacotherapy is a frequent problem in the disorder, few studies have systematically explored psychosocial factors related to treatment discontinuation. Previous research with depressed patients receiving psychotherapy has suggested that expectancies for improvement are related to treatment outcomes and that the therapeutic alliance may partially mediate this relationship. The current study found evidence for a similar relationship between patients' initial expectancies for improvement, patient and doctor-rated alliance, and long-term outcomes in bipolar patients treated with pharmacotherapy for up to 28 months following an acute episode. The results highlight the need for the assessment of expectancies and alliance in bipolar treatment and suggest possible targets for psychosocial interventions.