In each of two experiments, rats first received Pavlovian appetitive (food) conditioning in which white noise was established for different groups on the basis of either a two-or a four-pellet unconditioned stimulus (US) as a CS+ (a signal for the presence of food), CSo (a nonpredictive stimulus), or CS-(a signal for the absence of food). In Experiment 1, the conditioned stimulus (CS) was presented contingent on the food-reinforced response in a moderately difficult visual discrimination with the magnitude of the instrumental reinforcer matching that which the subject had received as a US. Relative to the performance of both the CSo and a novel-CS control, the CS+ retarded and the CS-facilitated learning for the two-pellet subjects but not for the four-pellet subjects. To eliminate any counteracting discriminability effects of the stronger four-pellet CSs, Experiment 2 employed the same design as Experiment 1 but instead used an easy visual discrimination. Again, the CS+ retarded and the CS-facilitated learning, but-as expected -these effects were now more pronounced for the four-than for the twopellet subjects. These findings are opposed to a general "cue" or signaling interpretation of the CS's properties. Instead, they offer strong support for a discrepancy (blocking) interpretation of both the within-and across-reinforcement (e.g., aversive to appetitive) transfer effects that are observed for CSs employed as response-contingent events proximal to a terminal reinforcer.Prior research investigating transfer from interpreted as reflecting dual-attribute learn-Pavlovian aversive (Av) to instramenal ing in Pavlovian conditioning: Depending appetitive (Ap) conditioning has shown on the correlation between the conditioned that a signal for shock (an AvCS+) con-stimulus (CS) and the unconditioned stimtingent on the food-reinforced response in ulus (US; cf. Rescorla, 1967), the primary a visual T-maze discrimination will facilitate associative property of the CS will be to learning, whereas a signal for the absence signal the presence (CS+) or absence of shock (an AvCS-) similarly applied (CS-) of a US; in addition, the CS will will retard such learning (Fowler, Fago, acquire a specific affective value reflecting Domber, & Hochhauser, 1973; Ghiselli & the nature of the US (i.e., Av or Ap) on Goodman & Fowler, 1976). which it is based. Assuming that the CS's These counterintuitive findings have been original affective value is a relatively transient property that is dependent on the US _,."~~~. 77', ~c 7 r presently available, it should be rapidly This study was submitted by the first author in v . r. . , . . ^ v ] partial fulfillment of the requirements for the extinguished and transformed to a new PhD degree at the University of Pittsburgh. The value in the context of a qualitatively differstudy was supported in part by United States ent reinforcer while leaving the CS's US-Public Health Service Grant MH-24115 to the s i gna ii ng va i ue largely intact. Hence, when Se