Delayed predictive inference integration with and revision by low-competitive inference alternatives in Chinese narrative text reading.

When readers encounter information conflicting with the predictive inferences made earlier, they may update the outdated ones with new ones, a process known as predictive inference revision. The current study examined the revision of disconfirmed predictive inferences by the primarily weakly activated, thus low-competitive inference alternatives during Chinese narrative text reading among Chinese native speakers. We conducted an event-related brain potential (ERP) experiment to study the predictive inference revision with increasingly supportive information for the low-competitive predictive inference alternatives. It serves as the very first attempts to study the predictive inference revision mechanisms by combining a larger range of ERP components, including frontal-Post-N400-Positivity (f-PNP) as an index of revision to examine the influences of the alternative inferences at later stages of reading comprehension. Our results showed that readers could detect inconsistent information (P300), disconfirm the incorrect predictive inferences before successfully integrating the low-competitive alternative predictive inferences with their current situation model (N400), engaging themselves in a second-pass reanalysis process incurring processing costs (P600), and revising the disconfirmed predictive inferences (f-PNP) at a later stage of reading comprehension. Results of this study are supportive of relevant theories in assuming that predictive inference revision does not happen immediately upon encountering conflicting information but happens slowly and incrementally. Our results also unfold the post-revision mechanisms by suggesting the remaining activation and lingering influences of the disconfirmed inferences in the forthcoming reading process.