Dimensional training attenuates overshadowing

Abstract

Prior experience with elemental or configural discriminations shapes how agents learn subsequent information: prior training that encourages elemental processing promotes competition between events, while prior configural training tends to attenuate competition. We sought to assess whether configural and elemental pretraining modulated subsequent overshadowing when using two separable dimensions (i.e., colors and symbols) in a predictive learning task. Furthermore, we used post-test questionnaires asking whether participants used either only one or both dimensions during training, and assessed if this modulated the magnitude of overshadowing. Across three experiments, prior experience with a configural discrimination (i.e., biconditional discrimination) did not modulate the magnitude of overshadowing compared to control groups. However, attended dimensions were less prone to overshadowing (Experiments 1 and 2), and Experiment 3 revealed that prior elemental training making one dimension relevant attenuated subsequent overshadowing of that dimension. Hence, preferred or trained dimensions are less prone to be overshadowed. Results are discussed in light of attentional and configural theories of associative learning.