A Meta-Analysis of the Reading Comprehension Skills of Students with Autism Spectrum Disorders

Background: Researchers have tended to suggest that students with ASD have weaknesses in reading comprehension despite superior decoding ability. Results have been conflicting, however, with some studies showing that students with ASD have typical reading comprehension and decoding skills, and others reporting impairments in both domains. To date, no formal meta-analysis has been conducted of reading comprehension skills in the ASD population.

Objectives: 1. Use meta-analysis to determine the size of the difference in reading comprehension between individuals with ASD and their typically developing (TD) peers as well as the direction and consistency of this effect. 2. To determine whether the effect sizes across studies are heterogeneous, i.e., whether there is true variation in effects. 3. Examine the degree to which decoding ability, vocabulary knowledge, Performance IQ (PIQ) and age are predictors for reading comprehension skill in the ASD population.

Methods: We conducted a meta-analysis of 33 studies of individuals with ASD that included a measure of reading comprehension. Effect sizes for the difference between the reading comprehension scores of individuals with ASD and their TD peers was calculated using Hedge’s g standardized mean difference. The statistical significance of an overall summary effect size was tested with a Z-test, and a Q-test was used to determine that the effects sizes for reading comprehension scores were heterogeneous. The four predictor variables (decoding, vocabulary, PIQ, age) were analyzed using the tau-squared method and the proportion of variance explained by each covariate was calculated.

Results: The summary effect for reading comprehension in individuals with ASD across all the studies was -1.0 SD. That is, individuals with ASD tended to score 1 SD below their TD peers in reading comprehension. However, this meta-analysis also found that the true effect size of the difference between groups does vary from one study to the next, varying between -2.3 SD to +0.3 SD. Moderator variables: Decoding, vocabulary knowledge and PIQ all significantly predicted the reading comprehension scores of individuals with ASD. Vocabulary was the strongest predictor of reading comprehension, explaining 90% of the variance in reading comprehension scores, followed by PIQ predicting 54% and decoding predicting 39%. Age was not a significant predictor of reading comprehension scores.

Conclusions: There was great variability in the reading comprehension scores of the individuals with ASD to the extent that some individuals had severe deficits in reading comprehension, whereas others had reading comprehension skills that were similar to their TD peers of the same age and grade level. It is important to note, however, that in many instances, individuals with ASD were compared to controls who had higher PIQs and/or greater vocabularies. Furthermore, some authors pre-selected individuals with ASD who had decoding strengths along with reading comprehension deficits and did not select individuals with different reading profiles. Overall, results suggest that having ASD increases the likelihood that a student may have problems with reading comprehension, but whether a given student actually has reading comprehension deficits may depend more on other factors (such as language ability) than simply their ASD diagnosis.