Making and revising predictive inferences during Chinese narrative text reading: Evidence from an electrophysiological study

The current study employed the event-related potential (ERP) technique to investigate predictive inference revision during Chinese narrative text reading among Chinese native speakers. Experiment 1 studied predictive inference revision by ensuring high contextual constraints for activation of the primary predictive inferences. Experiment 2 inspected the effects of the weaker inference alternatives on the revision process. Longer reading time and less positive mean average amplitude with two subcomponents of P300 (P3a and P3b) in the revise condition suggest that readers could detect inconsistent information and disconfirm the incorrect predictive inferences. However, they have difficulties in either integrating the alternative predictive inferences (N400) or revising the incorrect ones (P600), especially when the alternatives are of weaker activation levels. This study supports the Knowledge Revision Components (KReC) framework by verifying remaining activation of the disconfirmed primary inferences and extends it by considering effects of competitive alternatives on the predictive inference revision process.