Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Psycholinguistic constructs underlying words in naturalistic listening
Poster Session A - Sandbox Series, Thursday, October 24, 10:00 - 11:30 am, Great Hall 3 and 4
This poster is part of the Sandbox Series.
Xuan Yang1, Rutvik Desai1; 1University of South Carolina
Understanding how words are represented and organized in the brain is a crucial question in psycholinguistics and cognitive neuroscience. Most studies typically only consider a handful of lexical variables, such as frequency and concreteness. However, words have a large number of partially intercorrelated properties that influence their processing and neural organization. A more comprehensive assessment of the influence of these variables is crucial. Recent studies have begun investigating the neural representation of psycholinguistic variables in naturalistic contexts, such as listening to narratives in an MRI scanner. However, several methodological challenges limit the broader application of these studies, including the limited availability of myriad psycholinguistic variables in individual databases, and the lack of psycholinguistic values for all word forms in naturalistic stimuli. To address these issues, our study leverages the South Carolina Psycholinguistic metabase (SCOPE) (Gao et al., 2023), an extensive collection of psycholinguistic properties aggregated from major databases. By lemmatizing words, we maximize the number of tokens in narratives with available psycholinguistic values for subsequent data analysis. We used 113 psycholinguistic variables from SCOPE, spanning general, phonological, orthographic, semantic, and morphological categories. When original word form values were unavailable, lemma values were used, resulting in 12,375 unique words (5,876 unique lemmas) for analysis. We combined seven datasets from the publicly available Narratives fMRI database (Nastase et al., 2021), consisting of 213 fMRI scans from 102 neurotypical young adults listening to narratives in the scanner. Exploratory factor analysis was performed using the principal axis factoring extraction method and oblimin rotation to extract latent structures of psycholinguistic variables. Factor scores of single words were correlated with the fMRI data in separate models using parametric modulation regression. Six latent factors were extracted (explaining ~50% of the total variance), including those related to word frequency, word length, graphotactic/phonotactic probabilities, feedback consistency, feedforward consistency, and concreteness. These factors represent a total of 113 variables with loadings above 0.4. These factors significantly correlated with performance (response time and accuracy) in behavioral tasks such as lexical decision, word naming, and semantic decision. FMRI results showed that lower frequency factor was correlated with activation in IFG and MTG. Higher length was correlated with bilateral pSTG and SMG. Higher graphotactic/phonotactic probabilities were correlated with activation in bilateral STG. Lower feedback consistency factor was correlated with activation in left ATL. Higher feedback consistency was correlated with activation in bilateral posterior temporal lobe and IFG. Lower concreteness factor was correlated with activation in bilateral ATL and STG. The brain regions identified by these latent factors were largely consistent with previous findings using highly controlled stimuli and tasks. Novel result include spelling-sound consistency effects in a purely auditory task. This study explores the latent structures of psycholinguistic variables for 12,375 words, revealing six interpretable factors at both behavioral and neural levels. This approach provides a novel, comprehensive view of word representation in the brain, bridging the gap between isolated linguistic properties and potential underlying constructs. The findings have significant implications for our understanding of lexical processing in naturalistic contexts.
Topic Areas: Meaning: Lexical Semantics, Meaning: Discourse and Pragmatics