| Summary: | Five experiments with rats were conducted to determine the extent to which associative processes could be used to explain how rats seem able to learn complex CS-US contingencies during Pavlovian conditioning. Rats were exposed to positive, zero and negative CS-US contingencies and conditioned behaviour was compared with predictions derived from both associative models of conditioning and nonassociative normative theories of causal reasoning. A common measure of contingency, Deltap, when used to analyze Pavlovian conditioning requires defining the likelihood of the US in the presence and absence of the CS. Experiments 1 and 2 involved a novel preparation in which, in addition to standard CS presence trials, the absence of the CS was signalled by a second CS, called the trial marker (a lever). Rats were trained to learn relationships in which the CS was either a positive predictor of the US or in which it was unrelated to the US. More conditioned tray entries were observed when the CS signalled an increased likelihood of the US (positive contingency). Consistent with the associative explanations, the trial marker elicited conditioned lever pressing when the CS signalled no change in the likelihood of the US (zero contingency). Experiments 3, 4 and 5 extended the analysis with multiple CSs. These experiments examined whether learning about one CS was determined by its contingency relative to the contingency of other concurrently trained CSs. In experiments 3 and 4 conditioned responding to a moderately predictive CS was determined by its contingency relative to a perfectly predictive CS. Experiment 5 extended this effect to a case in which conditioning was influenced by the presence of a perfect predictor of the absence of the US. Together these results support the hypothesis that relative contingencies determine the strength of conditioned responding. The results are discussed from the perspective of both associative and nonassociative theory.
|
|---|
