Twenty boys (6-8 years) rated by their teachers as hyperactive and a matched sample of nonhyperactive boys performed a task that required them to withhold responding for a set time interval in order to be rewarded (DRL 6-second schedule). Half of each group worked on a one-button console while the other half was provided with additional collateral buttons. Results indicated that hyperactive children were relatively unable to perform efficiently on the task, and that this deficit endured regardless of age, IQ, or experimental condition. DRL was thus found to discriminate accurately between teacher-rated and parent-rated hyperactive and nonhyperactive children. Furthermore, a wide variety of self-generated mediating behaviors was observed, and it was determined that a child's DRL performance was related to the kind of mediating behaviors he displayed. Results are discussed in terms of the clinical assessment of hyperactivity and the training of impulsive children.