Résumé : Unipolar and bipolar patients with a chronic illness pattern were investigated to evaluate the relevance of clinical and psychosocial risk factors to predict subsequent recurrence. Self-esteem, social adjustment, social support and attributional style were assessed in 27 recovered bipolar patients, 24 recovered unipolar patients maintained on lithium or antidepressant prophylaxis and 26 healthy controls. They were further interviewed every 2 months in a 1-year period in order to diagnose affective episodes according to Research Diagnostic Criteria. Survival analyses and Cox's regressions demonstrated that being a unipolar patient and showing poor social adjustment were the strongest predictors of the occurrence of affective episodes. Self-esteem, social support, attributional style and clinical characteristics, such as age at illness onset, number of previous episodes or of previous hospitalizations and presence of affective disorder in first-degree relatives, were not found to be risk factors for further recurrence. This study stresses the importance of social adjustment in evaluating the outcome of affectively ill patients maintained on medication prophylaxis.