Content-embedded tasks beat complex span for predicting comprehension.

Discourse comprehension requires one to process information that is actively maintained in working memory (WM). Therefore, we hypothesized that individual differences in comprehension would be predicted better by working memory tasks that capture the concurrent demands of processing and maintenance of the same memory elements (i.e., content-embedded tasks) than by WM tasks that require the maintenance of an extraneous memory load during processing (e.g., complex span tasks). Two hundred sixty-one undergraduates completed three content-embedded tasks, three complex span tasks, and three measures of comprehension. Results of structural equation modeling indicated that the content-embedded tasks accounted for a greater amount of variance in comprehension than did complex span tasks. Thus, tasks that require one to coordinate the processing and maintenance of task-specific memory elements are preferable for capturing the relationship between WM and comprehension.