Decoding meaning composition during naturalistic language comprehension

Keywords

Loading...
Thumbnail Image

Issue Date

2022-06-22

Language

en

Document type

Journal Title

Journal ISSN

Volume Title

Publisher

Title

ISSN

Volume

Issue

Startpage

Endpage

DOI

Abstract

Human language allows for the combination of individual word meanings into more complex expressions. This capacity, meaning composition, and its neural bases have been studied typically using tightly controlled stimuli. The current study aimed to extend this body of research, asking if and to what extent findings from studies deploying tightly controlled stimuli generalise to more naturalistic contexts. Twenty-four participants listened to 45 minutes of audiobooks annotated for word onset, part of speech, and dependency relations while magnetoencephalography (MEG) was recorded. We used a series of decoding analyses to examine the multivariate neural responses to composition-related features. Our analyses were designed to isolate composition-related operations and representations whilst controlling for possible confounds such as linear word order, word distance, word position, and part of speech. We show that (1) composition-related features were uniquely reflected in the low frequency content of neural readout and (2) the coding of compositional features is in part time stable and in part time varying. Overall, our work shows that combining analysis techniques from engineering with temporally resolved neuroimaging grounded in theory yields new insights into the brain bases of meaning composition.

Description

Citation

Faculty

Faculteit der Sociale Wetenschappen