An experiment in which a pigeon was trained in contextual discrimination of its own behavior was carried out. When the experimental chamber was illuminated with a constant light, the pigeon had to peck on a red (or green) key in the sample component after having been pecking to the left (or to the right). When the chamber was illuminated with an intermittent light, the reinforced sample-comparison sequences were the opposite. The subject learned the task in about 40 sessions and maintained high correct response ratios even though the reinforcement probability decreased from 1 to 0.2 after each correct trial. The results are discussed in terms of the kind of discriminative rule and the kind of hierarchic structure involved in the task.