A data-driven approach to understanding reading comprehension difficulties
Abstract
OBJECTIVES: Some children struggle with reading comprehension despite success in decoding written words. Several studies indicate that oral language weaknesses play a causal role in poor comprehenders’ difficulties, yet small sample sizes and strict case-control comparisons are unable to capture the likely heterogeneity. We took a data-driven approach to identifying reading profiles in a large sample from the Avon Longitudinal Study of Parents and Children (ALSPAC), and examined whether varied cognitive profiles might underpin poor reading comprehension.
METHODS: Analysis plans were preregistered (https://osf.io/4zahf). We used latent mixture modelling to identify profiles of readers based on decoding and comprehension assessments administered at age 8-9 years (n = 6,846). With the children identified as poor comprehenders, a second mixture model included measures of vocabulary, nonverbal ability, working memory, and attention, to assess whether reading comprehension difficulties comprise several distinct cognitive profiles.
RESULTS: The preregistered model did not succeed in identifying varied profiles of decoding and comprehension skills. However, by controlling for general ability, a 6-class model emerged that incorporated a group with poor comprehension relative to their decoding skills (n = 947). These poor comprehenders had weak vocabulary and nonverbal ability relative to the rest of the sample. A small subgroup (18.9%) had broader cognitive difficulties, accompanied by more severe reading problems.
CONCLUSIONS: Reading comprehension difficulties are best considered along a continuum, with strengths and weaknesses in other cognitive domains likely influencing severity. Our ongoing research will document the educational and professional outcomes of poor comprehenders.
