Self-injury is a significant problem for many individuals with developmental disabilities, particularly those with severe or profound mental retardation. Many hypotheses have been suggested to account for self-injury, but none has been comprehensive. In this paper hypotheses suggesting psychological, behavioral, physiological, or neurochemical factors as causes of self-injury were critically reviewed. A compulsive behavior hypothesis was then introduced, which allows for alternative interpretations of some existing data and suggests several readily testable predictions.