College students responded under a multiple differential‐reinforcement‐of‐low‐rate 5‐s fixed‐ratio 8 schedule, with components alternating every 2 min. After 40 programmed minutes of acquisition and 12 min of maintenance, without notice, both schedules changed to extinction for 28 min. During acquisition, between alternations of the multiple schedule, some subjects were asked to develop rules describing the schedule contingencies. Other subjects were given these same rules between alternations, and a third group neither received nor were asked to develop rules. By the end of the acquisition phase, self‐generated‐rule subjects were more likely to show schedule‐typical behavior than were subjects not asked to generate rules. The behavior of those given rules was similar to those asked to generate rules at the end of acquisition, but yoked‐rule subjects acquired schedule‐typical behavior at a quicker rate. By the end of extinction, during the period corresponding to the previous fixed‐ratio interval, all no‐rule subjects who had earned points during acquisition and maintenance were responding at a rate of less than 30 responses per minute. Only 3 of the 9 self‐generated‐rule subjects and 2 of the 5 yoked‐rule subjects were similarly responding at this low rate. Results suggest that asking subjects to develop self‐rules facilitates acquisition, but can retard extinction. Results also suggest that self‐generated rules function similarly to external rules.