Individual typological differences in a neurally distributed semantic processing system: Revisiting the Science article by Mitchell et al. on computational neurolinguistics

Hiroyuki Akama

doi:10.12688/f1000research.14584.1

Abstract

Background: Revisiting the 2008 Science article by Mitchell et al. on computational neurolinguistics, individual typological differences were found as striking characteristics in the patterns of informative voxels crucial for the distributed semantic processing system. Methods: The results of different feature selection methods (ANOVA and Stability) were compared based on the open datasets of each subject for evaluating how these features were decisive in predicting human brain activity associated with language meaning. Results: In general, the two selection results were similar and the voxel-wise ranks were correlated but they became extremely dispersive for a subgroup of subjects exhibiting mediocre precision when examined without regularization. Quite interestingly, looking at the anatomical location of these voxels, it appears that the modality-specific areas were likely to be monitored by the Stability score (indexing “identity”), and that the ANOVA (emphasizing “difference”) tended to detect supramodal semantic areas. Conclusions: This minor finding indicates that in some cases, seemingly poor data may deeply and systematically conceal information that is significant and worthwhile. It may have potential for shedding new light on in the controversy pertaining to cognitive semantics, which is divided into modality-biased (embodied) and amodal symbol theories.

Highlights

It is widely acknowledged that despite some challenges in multivoxel pattern analysis (MVPA), the issue of individual variability, raised as a penalty to classification accuracy in cross-subject modelling, is challenging to overcome, when targeting concepts and meanings conveyed by language[1]
The raw modelling accuracy, based on the ordinary least square methods (OLS) without L2-regularization, decayed with the magnitude of “divergence” between the ANOVA and the Stability score and with the decrease in rank similarity between the voxels selected by each method (Table 1)
Our results support the notion that particular types of individuals differ markedly in their way of recruiting voxels with respect to different feature selection methods, i.e., Stability scores and F values of ANOVA

Summary

Introduction

It is widely acknowledged that despite some challenges in multivoxel pattern analysis (MVPA), the issue of individual variability, raised as a penalty to classification accuracy in cross-subject modelling, is challenging to overcome, when targeting concepts and meanings conveyed by language[1]. The precision rates in MVPA could be mostly uniform for experiments successfully performed at the individual first level, as in the case of the classically recognized Science article authored by T. Mitchell and his group (Predicting Human Brain Activity Associated with the Meanings of Nouns)[2]. Conclusions: This minor finding indicates that in some cases, seemingly poor data may deeply and systematically conceal information that is significant and worthwhile It may have potential for shedding new light on in the controversy pertaining to cognitive semantics, which is divided into modality-biased (embodied) and amodal symbol theories

Objectives

Methods

Results

Conclusion