Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Reading Skill Predicts Variability in Pseudoword Pronunciation: An Experience-Dependent Basis for Neuroimaging Analyses
Poster Session B, Friday, September 12, 4:30 - 6:00 pm, Field House
Michaela Brooks1, Caleb Solomon1, Robert Wiley2, Cory McCabe3, Lucia Zepeda-Rivera1, Donald Bolger1, William Graves3, Jeremy Purcell1; 1University of Maryland College Park, 2University of North Carolina at Greensboro, 3Rutgers University
In English, inconsistencies in spelling-to-sound mappings often yield multiple plausible pronunciations for a single pseudoword. While there have been neuroimaging studies investigating pseudoword reading and its relationship to reading skill, these studies are often limited by rigid binary definitions of pseudoword “accuracy” and overlook the variability in how pseudowords are actually pronounced. We propose that capturing this variability can offer a valuable behavioral anchor for interpreting individual differences in brain activation patterns during reading, consistent with experience-dependent and statistical learning frameworks. Fifty-one participants completed the Word Identification and Word Attack subtests of the Woodcock Reading Mastery Test (WRMT). Age-adjusted Basic Skills Composite scores were used to divide participants into three reading skill groups. During an fMRI scan, participants read aloud 40 pseudowords; responses were recorded and phonetically transcribed. To better characterize pronunciation variability, we defined several accuracy thresholds based on sample-level response distributions of pronunciations where accurate was defined the following six ways starting with the most stringent and proceeding to the most lenient: (1) most frequently produced in the sample; (2) among the top two most frequently produced in the sample; (3-5) in the top 25%, 50%, or 75% of cumulative response probabilities, and (6) produced by at least one other participant. For each participant, the proportion of accurate pronunciations under each threshold was calculated and correlated with WRMT reading scores. All within-sample accuracy definitions showed significant positive correlations with reading skill (all ps < .05*), with the strongest association observed for the top 75% cumulative response (r = 0.54, p < 0.001***). To assess how pronunciation variability differed by reading skill group, we calculated the mean Shannon entropy for each pseudoword spelling across groups. A linear mixed-effects model predicting entropy from reading group (with random intercepts for pseudowords) revealed a significant effect of reading skill group. Specifically, compared to the high-skilled reading group, both the average- and low-skilled groups demonstrated significantly higher entropy values (Average: t(78) = 2.23, p = 0.02*; Low: t(78) = 9.12, p < 0.001***). These findings demonstrate that readers with higher skill levels produce more consistent pronunciations of pseudowords, as indexed by group-based probability thresholds and entropy. Capturing this behavioral variability may provide a foundation to interpret more granular individual differences in neural responses during decoding, moving beyond binary notions of accuracy for pseudowords, and enhance our understanding of sublexical processing.
Topic Areas: Reading, Writing and Spelling