Nonce word evidence for the misinterpretation of implausible events

Good-Enough Processing accounts posit a two-stream mechanism by which an algorithmic, bottom-up parse is simultaneously built alongside a heuristic, top-down parse that is prone, in real-time, to influences from real-world expectations, which sometimes leads to misinterpretations of implausible events. Post-interpretive accounts suggest the offline findings often used as evidence introduce interference due to the memory they require, favouring instead an algorithmic-only account of parsing. The current study uses self-paced reading, question answering, and sentence completions to provide converging evidence for these misinterpretations, using nonce-nouns as a baseline for increased working memory burden against which event plausibility can be compared. The findings reveal a pattern where implausible sentences rarely cause online processing difficulty compared to plausible sentences while at the same time resulting in higher rates of misinterpretation. The data favour a Good-Enough processing account and highlight the issues with relying solely on online methods for psycholinguistic inquiry.