FN400 amplitudes reveal the differentiation of semantic inferences within natural vs. artificial domains

Abstract

Category-based inferences allow inductions about novel properties based on categorical memberships (e.g., knowing all trout have genes [premise] allows us to infer that all fish have genes [conclusion]). Natural (N) and artificial (A) domains are the most obvious and traditional distinctions in categorization. The distinct event-related potential (ERP) responses for N and A domains have not yet been examined during category-based inferences. In this study, the differences between ERP inference parameters within N and A domains were measured during inductive decision processing, while controlling the premise-conclusion similarity and premise typicality between those two domains. Twenty-two adults were asked to make a decision on whether a conclusion was definitely weak, possibly weak, possibly strong, or definitely strong, based on a premise. The behavioral results showed that semantic inferences within the N domain shared similar inductive strength, similar "correct" response rates, and similar reaction times with that within the A domain. However, the ERP results showed that semantic inferences elicited smaller frontal-distributed N400 (FN400) amplitudes within the N domain than within the A domain, which suggested that knowledge of the ontological domain of a category affects category-based inferences, and underlaid the increased categorical coherence and homogeneity in the N as compared to the A categories. Therefore, we have distinguished the cognitive course of semantic inferences between N and A domains.

Figure 1

Experimental procedure.

Figure 2

The electrode layout and the region of interest (ROI). The red cluster is ROI that be used for the further two-factor repeated-measures analyses of variances (ANOVAs) for ERP data.

Figure 3

The averaged inference strength, “correct” response rates (ACC), and reaction times (RTs) of “correct” responses for each sub-condition. N+ indicates “congruent conclusions with a natural premise category”; A+ indicates “congruent conclusions with an artificial premise category”; N− indicates “incongruent conclusions with a natural premise category”; A− indicates “incongruent conclusions with an artificial premise category”. Error bars represent mean ± s.e.m. ***p < 0.001.

Figure 4

The ERP responses to the congruency effects. Figure 4a shows the grandaveraged waveforms elicited by N+ and N− arguments and the difference waveform (N− minus N+) in the fronto-central region, and the topography of the difference waveform at 250–400-ms. Figure 4b illustrates the grand-averaged waveforms elicited by A+ and A− arguments and the difference waveform (A− minus A+) in the fronto-central region, and the topography of the difference waveform at 250–400-ms. N+ indicates “congruent conclusions with a natural premise category”; A+ indicates “congruent conclusions with an artificial premise category”; N− indicates “incongruent conclusions with a natural premise category”; A− indicates “incongruent conclusions with an artificial premise category”.

Figure 5

The ERP responses to the domain effects. Figure 5a shows the grandaveraged waveforms elicited by N+ and A+ arguments and the difference waveform (A+ minus N+) in the fronto-central region, and the topography of the difference waveform at 250–400-ms. Figure 5b illustrates the grand-averaged waveforms elicited by N− and A− arguments and the difference waveform (A− minus N−) in the fronto-central region, and the topography of the difference waveform at 250–400-ms. N+ indicates “congruent conclusions with a natural premise category”; A+ indicates “congruent conclusions with an artificial premise category”; N− indicates “incongruent conclusions with a natural premise category”; A− indicates “incongruent conclusions with an artificial premise category”.