A significant proportion of American high school students struggle with reading comprehension. Several different models might help identify the components that have the largest effect on comprehension. The current dissertation study replicates a comparison of the Construction-Integration (CI), Verbal Efficiency (VE), and Inferential Mediation (IM) models of reading comprehension, the latter model based on an extensive literature review. It then tests the fit of four variations on the IM model.

Ninth-grade students ranging from 1st to 99th percentile on comprehension completed measures of background knowledge, inferencing, strategies, vocabulary, word reading and comprehension. Researcher-developed measures of background knowledge, inferencing and strategies (based on Cromley & Azevedo, 2004a) showed good reliability with this sample.

A subset of the students also completed a think-aloud protocol while reading a passage from an American history textbook. These protocols were transcribed and coded using a coding scheme adapted from Azevedo, Guthrie, and Seibert (2004).

As in a preliminary study, the IM model had a much better fit to the data than did the CI or VE models. The original IM Model had the best fit, explaining 66% of the variance in comprehension. All predictors made a significant contribution to comprehension, with vocabulary, background knowledge, and strategies having significant indirect effects. Vocabulary and background knowledge made the greatest total contribution to comprehension. There were large, significant differences between low- and high-comprehending participants on all of the predictor variables, except for word reading accuracy, where there were small but significant differences.

The coded think-aloud protocols were largely consistent with the correlations underlying the model. Spearman rank correlations among the codes provide convergent evidence for eleven of the correlations underlying the model. The think-aloud protocols also provided convergent evidence for the validity of the paper-and-pencil measures. 

The current study validates and refines a new model of reading comprehension. Results suggest that both the direct and indirect effects of the components are important for comprehension. Results also suggest that vocabulary and background knowledge might first be targeted for interventions with 9th grade students who struggle with reading comprehension. Implications for future research are also discussed.