Rats in two separate experiments were trained on schedules of reinforcement that allowed the probability of reinforcement (a drop of water) given a leverpress in any second and the probability of water given no leverpress in any second to be manipulated independently of one another. After training on high-positive contingencies (no water for not leverpressing), rats were shifted to zero contingency (equal water for pressing or not pressing), and, as is usually found, 1ever-pressing declined radically. Then each water reinforcer given for not leverpressing was signaled by a 5-sec flashing light and gradually, after many sessions, leverpressing returned to normal. This effect was seen for all five experimental rats in a within-subject-design experiment and also for all four experimental rats-but not at all for the four control rats-in a second, between-subjectdesign experiment. These results are interpreted within a framework that stresses three-term relations between S, R, and reinforcement that emphasizes the difference between earned and unearned reinforcement as two of those three terms.In both classical and operant conditioning, the presentation of a substantial number of USs or reinforcers that are not preceded by a response or a punctate stimulus can result in substantial loss of conditioning (e.g., Rescorla, 1968;Rescorla & Skucy, 1969). These poor conditioning situations are often described by the zero contingency or correlation between the two elements in conditioningstimulus and US or response and reinforcer (Rescorla, 1967).In classical conditioning, some theories, such as the Rescorla-Wagner model (Rescorla & Wagner, 1972), have emphasized the role that conditioning of the context plays in mediating such effects. According to this model, unsignaled USs condition the context or background cues, but when those USs are signaled by an entirely different CS, competition occurs between this new CS and the context. This view predicts that such added CSs before USs during the intertrial interval will make the zero contingency ineffective and that conditioning will now occur to the original CS. Most, but not all, experiments that have compared the effect of signaled versus unsignaled intertrial USs have supported this model. Durlach (1983) reports two such experiments and reviews many other such studies. In the classical conditioning literature, it can also be noted that there is a tangentially related effect of signaling intertrial USs upon the development or the maintenance of inhibition (Baker, 1977;Hammond, 1966Hammond, , 1968.The authors thank Philip Hineline for editorial assistance in the preparation of the manuscript. This research was supported in part by a grantin-aid from Temple University. Reprints may be obtained from the senior author at the Department of Psychology, Temple University, Philadelphia, PA 19122. In operant conditioning, one can also arrange for a zero correlation or contingency between the elements of conditioning, in this case between response and reinforcement (RFT), and observ...