How antecedent retrieval influences prediction: a cue-based retrieval model

AMLaP 2020, Poster ID: 241, Poster session 2 (Friday, 4 September 2020) Stone et al. (2020) reported visual-world eye tracking data showing that predictions of the upcoming possessum were faster when the possessum and possessor matched in gender than when they mismatched. This prediction advantage was reflected in the target fixation probabilities. The effect is not expected under existing theories of prediction because the gender of the possessor doesn't provide any extra information about the upcoming possessum. Here we explain the effect through a computational model in the cue-based retrieval framework. The model captures the effect through interaction between interference in the antecedent retrieval process and the target prediction process. Consequently, we also propose an ACT-R activation-based model of fixation probabilities.