Although polysomnographic abnormalities are prevalent in adults with major depressive disorders (MDD), the findings in children and adolescents have been more equivocal. Polysomnographic measures may be of predictive value in assessing course of illness. The present study used standard sleep measures and temporal coherence of sleep electroencephalogram (EEG) rhythms to predict recovery and recurrence in a 1-yr naturalistic follow-up in 47 children and adolescents 8-18 yr of age with MDD. Standard sleep measures did not predict clinical course. On the other hand, temporal coherence measures discriminated between those who recovered, recovered but recurred, and those who did not recover from the index episode. Specifically, coherence between beta, theta and delta recorded in the right hemisphere was significantly lower in the no-recovery group. In addition, temporal coherence was strongly associated with both time to recovery and recurrence. Those with the lowest coherence were less likely to recover or recurred sooner. Significant sex differences were found with a stronger relationship between temporal coherence and clinical course in boys. This study supports the use of quantitative sleep EEG measures as a predictor of clinical course in depression.