Relational integration predicted numerical inductive reasoning: ERP Evidence from the N400 and LNC.

Abstract

As relational integration performance can be used to predict reasoning ability, the present study aimed to provide electrophysiological evidence for numerical inductive reasoning. Number series with two levels of relational complexity were utilized, including simple and hierarchical problems (such as "15-16-17" versus "15-16-18"). Two tasks were adopted: a relational integration task that required to determine whether the numerical relations were changed across numbers; a number series task that required to determine whether a hidden rule was acquired (Experiment 1) or to predict the subsequent number (Experiment 2), whose phases were divided as rule searching, rule discovery, and rule following. The event-related potential (ERP) results of both experiments indicated that, in contrast to simple problems, hierarchical problems triggered enhanced N400 and late negative component (LNC), reflecting numerical fact retrieval, and generalizing novel hypotheses about the hidden rules by integrating adjacent numerical relations, respectively; relational integration showed similar N400 and LNC activation patterns to rule discovery (Experiment 1) or rule searching (Experiment 2). Additionally, the N400 and LNC elicited by relational integration showed strong positive correlations and even were able to predict the ones triggered by rule discovery (Experiment 1) or rule searching (Experiment 2). Therefore, the results supported the role of relational integration in numerical inductive reasoning and thereby in intelligence.