Representational shifts as a distinct mechanism in associative learning and generalization

The peak shift effect in generalization, a robust phenomenon where the maximal response occurs to stimuli shifted away from the conditioned stimulus in the direction opposite the inhibitory stimulus, has been predominantly attributed to associative learning. However, recent findings point to biases in stimulus representation as an alternative mechanism. Our research, comprising two experiments (Experiment 1: N = 76; Experiment 2: N = 234), investigated whether the peak shift effect could emerge from a stimulus contrast effect, independent of associative learning. By comparing stimulus identification patterns between simple and differential conditioning groups both after stimulus exposure and following associative learning, we found that shifts in the identification of conditioned stimuli were primarily driven by a stimulus contrast effect, rather than associative learning per se. This representational shift was closely associated with the peak shift phenomenon observed in post-learning generalization patterns, with identification errors strongly impacting generalized responding. These results underscore the importance of considering stimulus representations as an independent process in understanding learning-based behavior, offering new insights into the mechanisms underlying generalization phenomena.