Using Construction-Integration Theory to Interpret Reading Comprehension Instruction for Students With Autism Spectrum Disorder: A Systematic Review and Meta-Analysis
Reading Research Quarterly
Zhang, S., Prykanowski, D. A., et al. (2023).
Reading Research Quarterly, 58(1), 126-159.
This systematic review investigates the effects of intervention on reading comprehension for students on the autism spectrum.
Not stated
From database inception to March 2020
Empirically-based experimental studies
26
The immediate effect of reading comprehension interventions for single-subject research designs (SSRDs) was small (Tau-U = 0.62, 95% CI = [0.44, 0.80]) and the maintenance effect was medium (Tau-U = 0.81, 95% CI = [0.66, 1.00]). For group designs, small but significant effects were found for reading comprehension (g = 0.26, 95% CI = [0.13, 0.39]). Immediate effects (g = 0.29, 95% CI = [0.13, 0.46]) were larger than maintenance effects (g = 0.15, 95% CI = [0.07, 0.24]). Most studies had relatively high design quality, but few studies demonstrated functional relations. SSRD studies showed greater risk of bias than group designs.
Subgroup analysis revealed potential advantages of a situation model level of instruction (i.e., information from the text integrated with the reader’s world knowledge). Tau-U: Situation = 0.74, Textbase = 0.72, Surface = 0.14; Hedges’s g: Situation = 0.31, Textbase = 0.25, Surface = 0.14. In SSRD studies, cloze scores were significantly lower than curriculum-based measure composite reading scores (cloze vs. composite b = −0.31, 95% CI = [−0.58, −0.05]). In group studies, curriculum-based retelling measures had better outcomes than norm-referenced reading measures (retelling vs. norm-referenced b = 1.34, 95% CI = [1.12, 1.46]).
For levels of text representation in SSRD studies, surface level (i.e., the actual words of the text) showed significantly lower effects than situation model level of instruction (surface vs. situation b = −0.61, 95% CI = [−0.84, −0.38]). Textbase (i.e., a set of text-based propositions and their relationships) and situation model did not differ as far as reading comprehension outcomes. In group studies, both surface level and the textbase level of processing had significantly smaller effects than instruction involving situation model level (surface vs. situation b = −0.19, 95% CI = [−0.32, −0.05]; textbase vs. situation b = −0.16, 95% CI = [−0.33, 0.00]).
