Regression analyses of event-related brain potentials dissociate general reading experience and domain-specific knowledge sources during language comprehension.

Abstract

How do different sources of knowledge and experience shape how we make meaning from language in real time? The present work capitalizes on recent advances in ERP analysis (the regression ERP, or rERP approach, combined with AIC model comparison to compare how various measures of knowledge and experience related to millisecond-by-millisecond ERP time-course data recorded as individuals read sentences about a fictional world (Harry Potter; HP) one word at a time. Participants were individuals who ranged in the experiences they had contributing to their HP knowledge, including the number of times they had read each book and/or seen each movie. These participants also completed an objective measure of domain knowledge using a short trivia quiz, as well as more general measures of reading experience. HP-based measures out-performed general reading experience for multiple indices of word understanding. Notably, self-reported experience with HP movie-watching best explained N400 "related anomaly" effects, which reflect the immediate availability of linguistically inappropriate but event-related words. Thus, the rich multimodal experiences of watching movies may, at least in some cases, accelerate knowledge use during language understanding compared to purely linguistic experiences of reading. These findings run counter to theories on which meaning is based on linguistic experience alone, implicating multimodal contributions. More generally, these analyses illustrate how exploratory model set comparisons and continuous variable rERPs yield new insights into the time course of information processing during reading.