The purpose of this article is to describe theoretical and research bases for constraint-induced movement therapy (CIMT), to discuss key features and variations in protocols currently in use with children, and to review the results of studies of efficacy. CIMT has been found to be an effective intervention for increasing functional use of the hemiparetic upper extremity in adults with chronic disability from stroke. CIMT developed out of behavioral research on the phenomenon of "learned nonuse" of an upper extremity, commonly observed following sensory and/or motor CNS injury, in which failure to regain use persists even after a period of partial recovery. CIMT includes three key elements: (1) constraining the use of the less-impaired upper extremity (UE); (2) intensive, repetitive daily therapist-directed practice of motor movements with the impaired UE for an extended period (2-3 weeks); and (3) shaping of more complex action patterns through a process of rewarding successive approximations to the target action. Mechanisms responsible for success are thought to be separate but complementary, that is, operant conditioning (reversal of learned nonuse) and experience-driven cortical reorganization. CIMT has recently been extended to children with hemiparesis secondary to perinatal stroke or other CNS pathology. Numerous case studies, as well as a small number of randomized controlled or controlled clinical trials have reported substantial gains in functional use of the hemiplegic UE following CIMT with children. Protocols vary widely in terms of type of constraint used, intensity and duration of training, and outcome measures. In general, all report gains in functional use, with minimal or no adverse effects. Continued research is needed, to clarify optimal protocol parameters and to further understand mechanisms of efficacy. (Contains 1 table.)