The concept of self-reinforcement was examined as an example of control of human behavior without direct, external reinforcement. Ss at 3 levels of incentive were given a learning period on a verbal discrimination task, with a light as reinforcer. In Phase n, they received 1 of 3 reinforcing procedures: SR (in which S was given complete control of administering reinforcing stimuli), continued acquistion, and extinction. The 3 procedures differed significantly in frequency of correct responses during Phase n, with the SR group lying between acquisition and extinction. The highest incentive group showed the greatest accuracy in taking available reinforcement; incentive did not affect total frequency of reinforcements taken.