Reinforcing one response class and withholding reinforcement for all other responses that are not part of that response class. We use differential reinforcement to reduce problem behavior by withholding all reinforcement.