This paper points to factors that determine whether repetitive stereotyped behavior occur in the behavior repertoire. The analysis pits an "intrinsic oscillator" mechanism against a "self-stimulation" theory and chooses to emphasize the latter. The paper accounts for the repetitive and rhythmic nature of stereotypy by suggesting that repetition in a rhythmic way is the most efficient way of self-stimulation. It proposes that rhythm may be a reinforcement in at least some cases. It raises the question of whether control of stimulation by the person is a necessary condition for maintaining stereotypy. The paper recognizes the possibility that stereotyped behaviors may have their origin in the common repetitive behaviors of infancy but emphasizes that pathological stereotypy may involve more than immature repetition. It suggests that there is reason to believe that early intervention to prevent pathological stereotyped behavior might be effective but that we do not know much about how stereotypies get started.