Non-literal Syntactic Representations are Driven by Predictive Processing: Evidence from Meta-Analysis and Speaker Accent

Abstract

Many sentences are grammatically ill-formed, contain speech errors, or are otherwise anomalous in some way. As a result, it is often reasonable for comprehenders to infer that implausible sentences may not reflect the true intent of the speaker, and thus interpret the sentence non-literally. Recent work (Cai, Zhao, & Pickering, 2022; Buxó-Lugo & Slevc, n.d.) suggests non-literal interpretation reflects syntactic representations, not just semantics. Cai et al. (2022) propose the plausibility-driven syntactic prediction account to explain how comprehenders may obtain non-veridical structural representations. Across two experiments, we attempt to replicate their findings and test the claim that structural prediction can lead to non-veridical representations. We use structural priming and comprehension questions to assess comprehenders’ structural representation and their interpretation of the sentence. We also present a small-scale meta-analysis of all studies to date using this paradigm. Our results provide support for the prediction account, by showing that when comprehenders rely more on prediction, they have a larger tendency to obtain non-veridical structural representations.