Ambiguity and context processing in human predictive learning.

Abstract

Two experiments explored the role of ambiguity on context processing by using relative stimulus validity designs in human predictive learning. Two groups of participants were trained with 2 stimulus compounds (XY and XZ). In Group TD (true discrimination), compound XY was always followed by the outcome, whereas compound XZ was never followed by it. In Group PD (pseudodiscrimination) the presentation of each compound was followed by the outcome in half of the trials. Experiment 1 found that pseudodiscrimination facilitated context dependency of reliable predictors regardless of whether they were trained in the same context in which pseudodiscrimination took place or in an alternative context in which true discrimination was conducted. Experiment 2 found context dependency of reliable predictors trained and tested in PD contexts, suggesting that the ambiguity in the meaning of the cues produced by pseudodiscrimination training is at least partially responsible for the context switch effects found in ambiguous situations in human predictive learning.