Recent research findings suggest that the initial reductive effects of noncontingent reinforcement (NCR) schedules on destructive behavior result from the establishing effects of an antecedent stimulus (i.e., the availability of "free" reinforcement) rather than extinction. A number of authors have suggested that these antecedent effects result primarily from reinforcer satiation, but an alternative hypothesis is that the individual attempts to access contingent reinforcement primarily when noncontingent reinforcement is unavailable, but chooses not to access contingent reinforcement when noncontingent reinforcement is available. If the satiation hypothesis is more accurate, then the reductive effects of NCR should increase over the course of a session, especially for denser schedules of NCR, and should occur during both NCR delivery and the NCR inter-reinforcement interval (NCR IRI). If the choice hypothesis is more accurate, then the reductive effects of NCR should be relatively constant over the course of a session for both denser and leaner schedules of NCR and should occur almost exclusively during the NCR interval (rather than the NCR IRI). To evaluate these hypotheses, we examined within-session trends of destructive behavior with denser and leaner schedules of NCR (without extinction), and also measured responding in the NCR interval separate from responding in the NCR IRI. Reductions in destructive behavior were mostly due to the participants choosing not to access contingent reinforcement when NCR was being delivered and only minimally due to reinforcer satiation.