Poster A44, Thursday, August 16, 10:15 am – 12:00 pm, Room 2000AB
Individual Differences in Text Comprehension: A Resting-State Functional Connectivity Study
Anya Yu1, Benjamin Schloss1, Chun-Ting Hsu1, Ping Li1;1Department of Psychology, The Pennsylvania State University
Reading is one of the fundamental methods through which we acquire new knowledge and skills, and is especially important in the absence of an immersed environment or an instructor (Hanh et al., 2007; Macabasco-O’Connell et al., 2011). Reading comprehension has been shown to be a strong predictor of individual’s quality of life as well as future success (Baker, Parker, Williams, Clark & Nurss, 1997), and individual characteristics such as executive function (EF) skills have been reported to be a significant factor that influences reading comprehension success in both children and adults (Cartwright, 2015). Text reading comprehension is a complex cognitive process relative to other processes, and relies on a distributed network of brain regions (Li & Clariana, 2018). It is therefore very likely that text reading comprehension is better captured by an interconnected and interactive neural network. Most neurocognitive investigations of language comprehension are limited to word-level rather than text-level reading (see reviews by Ferstl, 2010 and Mason & Just, 2013). Furthermore, most neuroimaging studies of text comprehension have been focused on investigating reading-related patterns via functional magnetic resonance imaging (fMRI), and evidence showing that resting-state functional connectivity (RSFC) can capture text reading comprehension is lacking. Our study aims to clarify the relationship between RSFC in the language network and reading comprehension performance (measured by Gray’s Silent Reading Task) as well as EF skills (measured by the Attention Network Test and Letter-Number Sequencing). Forty-six native English adults were recruited to read five expository scientific texts in the scanner. ROI peak selection was informed by literature on regions correlated with text-reading (see Koyama et al., 2010; Li & Clariana, 2018), which includes the left IFG, SMG, DLPFC, AG, and VWFA. A step-wise algorithm employing an OLS regression model was used to explore whether one or more two-way interactions could better explain variation in the GSRT scores. To address concerns about not adequately controlling for multiple comparisons and overfitting the data, we also used a model based on the decision regression tree algorithm (Breiman, 2001) that has been applied in functional connectivity studies (Richiardi, Eryilmaz, Schwartz, Vuilleumier & Van De Ville, 2010; Venkataraman, Whitford, Westin, Golland & Kubicki, 2012). All of the interactions that explained a significant amount of variance in the data are entered in a leave-one-subject-out cross-validation analysis. The behavioral results confirmed a significant positive correlation between executive function task performances and our reading task performance. While no single predictor had significant main effects with reading and EF indices, the decision tree model revealed significant effect in the VWFA-SMG and SMG-AG interaction that had above chance predicting power on GSRT performance (Spearman’s =.37, p=.01). These patterns suggest that the temporoparietal connectivity can act as a reliable classifier distinguishing relatively bad (mean standardized z-score <-5) and good (mean standardized z-score>1) readers. This is convergent with DTI finding correlating temporoparietal white-matter tract integrity with reading performance (Kingberg et al., 2000), suggesting this connectivity is particularly engaged in text comprehension.
Topic Area: Meaning: Discourse and Pragmatics