Multinomial processing tree models of recent accounts of the processing of noncanonical word order sentences

Abstract

This study investigates the predictions of the retrieval and the parsing accounts of the processing of noncanonical word order sentences by implementing their suggestions in multinomial processing tree (MPT) models (Riefer & Batchelder, 1988). MPT models allow the estimation of probabilities of unobservable and hypothesized cognitive events or states using categorical data, and an analysis of the predictions and the technical properties of these models can provide information as to how accounts of cognitive processes can be improved. Recent literature has identified systematic performance decrease in the agent/patient naming task, among other types of experimental tasks, when comprehenders were faced with noncanonical word order sentences (Ferreira, 2003; Bader & Meng, 2018). The retrieval account (Bader & Meng, 2018; Meng & Bader, 2021), suggests that this decrease in performance is caused by problems with the cue-based retrieval operation triggered by the task probes, whereas the parsing account (Ferreira, 2003; Christianson, Luke & Ferreira, 2010) suggests that the decrease in performance is caused by misinterpretation. We developed multiple MPT models that reflect the assumptions of these two accounts and fitted these to the data from the first experiment of Meng and Bader (2021). We found that the retrieval account is more eligible for adaptation into an MPT structure than the parsing account, and that the parsing account needed deeper revision of its assumptions.