If a response key is regularly illuminated for several seconds before food is presented, pigeons will peck it after a moderate number of pairings; this "auto-shaping" procedure of Brown and Jenkins (1968) was explored further in the present series of four experiments. The first showed that pecking was maintained even when pecks turned off the key and prevented reinforcement (auto-maintenance); the second controlled for possible effects of generalization and stimulus change. Two other experiments explored procedures that manipulated the tendency to peck the negatively correlated key by introducing alternative response keys which had no scheduled consequences. The results indicate that pecking can be established and maintained by certain stimulus-reinforcer relationships, independent of explicit or adventitious contingencies between response and reinforcer.