Investigating associations between executive function and reading comprehension in children with and without specific reading comprehension deficits

Investigating associations between executive function and reading comprehension in children with and without specific reading comprehension deficits

First Author: Mercedes Spencer -- Vanderbilt University
Additional authors/chairs: 
Laurie E. Cutting
Keywords: Reading comprehension, Poor Comprehenders, Executive Function, Structural equation modelling
Abstract / Summary: 

Purpose: The aim of the current study was to further investigate the unique contribution of executive function (EF) to reading comprehension (RC) by examining relations between EF, language comprehension (LC), decoding, and RC within a structural equation modeling framework. More specifically, we aimed to identify whether relations between skills differed for children with specific RC deficits (S-RCD) versus typical readers.
Method: Participants included 167 children (7.33–9.33 years old; 55.7% female). We measured EF, decoding, LC, and RC using a rich battery of assessments and created composite scores for each construct. Outliers were replaced with values at the high or low end of the range, and missing data was handled using maximum likelihood estimation. We used path analysis to test relations between EF, LC, decoding, and RC for the full sample and for children with S-RCD (N = 25; scores ≥ the 40th percentile on at least two measures of decoding and ≤ the 30th percentile on at least one measure of RC) and typical readers (N = 81; scores ≥ the 40th percentile on at least two measures of decoding and ≥ the 40th percentile on both measures of RC). We also conducted multi-group modeling to investigate potential differences in relations.
Results: For the full sample, decoding, LC, and EF all accounted for unique variance in RC. When children with S-RCD and typical readers were modeled separately, however, differences emerged. For children with S-RCD, decoding and EF continued to predict unique variance in RC but LC no longer accounted for additional variance. For typical readers, LC and decoding accounted for unique variance in RC but EF did not. Multi-group modeling also indicated that associations between LC and RC may vary across readers.
Conclusions: These findings support previous work showing that relations between EF and RC may be stronger for struggling readers and highlight the need to consider EF within theoretical frameworks of RC.