Validation of the direct and inferential mediation (DIME) model of reading comprehension in grades 7 through 12

Abstract This study evaluated the direct and inferential mediation (DIME) model of reading comprehension (Cromley & Azevedo, 2007) in a large ( n  = 1196) and diverse (grades 7–12) sample. Multi-indicator latent variables were used to measure six primary constructs: knowledge, vocabulary, word reading, strategies, inference, and reading comprehension. Results corroborated prior research when similar methods were used, but departed from prior findings when measurement error and shared method variance between the predictors and the reading comprehension outcomes (method bias) were controlled. Results generalized across middle and high school, and component skills of reading accounted for virtually all of the systematic variance in reading comprehension. Importantly, controlling method bias diminished the importance of knowledge and vocabulary, and increased the importance of inferencing. Mediated effects of knowledge and vocabulary through inference making were also found. The present study provided a stronger and more generalizable formulation of the DIME model than prior research, and highlighted limitations of using reading-based measures of predictors in component skills models of reading comprehension.