Multiple automatic and controlled working memory processes in comprehension: A latent variable approach

Several memory processes have been proposed to account for individual differences in discourse comprehension. Early research in reading comprehension focused on the correlation between reading comprehension and Baddeley's multiple components model of working memory (WM). More recently, a number of theorists have stressed that the empirically indicated limits of attention-driven WM storage components are inconsistent with the amounts of information needed to accomplish complex cognitive tasks, including comprehension. Several theorists have proposed models of WM that include available long-term memory (ALTM) as part of the cognitive workspace. ALTM refers to elements of long-term memory (LTM) whose resting strengths have been temporarily increased, but not to the point of conscious awareness. Others have argued that the ability to inhibit irrelevant information is the source of individual differences in comprehension, though there is much disagreement as to whether inhibitory processes are passive or active in nature. Passive deactivation of irrelevant information and the active disengagement of attention were operationalized as separate memory processes in the current study. The study used recently created measures of ALTM, deactivation, content specific background knowledge, and listening comprehension, as well as traditional measures of WM and Stroop interference. Structural equation modeling (SEM) of the relationships among factors representing these measures suggested several conclusions. First, measures of ATLM and deactivation capture similar passive memory processes. This finding suggests that the activation and deactivation of LTM elements may rely on the same cognitive resources. Second, the relationship between ALTM processes and listening comprehension was mediated by differences in amount of content specific knowledge. This finding supports theories of language comprehension that include the activation of background knowledge as a key component of comprehension. Third, the relationship between WM and listening comprehension was mediated by ALTM. This mediated relationship reflects how attention focus in WM results in greater availability of related LTM structures that more directly support the comprehension processes. In total, modeling of individual differences in these constructs offers a resolution to the apparent contradictions posed by limited capacity estimates of attention-driven WM components and comprehension theories that describe the integration of text information and background knowledge.