The activation and long-term memory of predictive inferences: The role of working memory constraint and text elaboration

The study was focused on the effects of working memory (WM) capacity and text elaboration on the processing and product of text comprehension. Participants included 73 English-native-speaking university students. A 2 (inferential vs. control target sentence) X 2 (low vs. high elaboration) X 2 (low- vs. high-span) mixed factorial design was conducted, with the former two variables as the within-subject external text factors, and the latter one as the between-subject inter-individual difference factor. Based upon the top and bottom third of a tripartite split of the participants. Read Span (RSPAN; Engle, 2005) percentile scores, 22 low- and 25 high-working memory WM span readers were identified. The activation of primary predictive inferences was measured by the naming task. The text memory was measured by the cued recall task. The reading materials were 24 narrative stories. In each story, the target sentence contained either the inference-evoking or control context for one primary predictive inference. The previous context of each story had either a low or high elaborative causal relation with the event described by the target sentence, thus eliciting either a low or high level of activation of a secondary inference.;The results demonstrated that difference mechanisms seem to be working for readers with different WM capacities and reading skills in terms of influencing their capability of the online inference computation. In addition, the working memory capacity contributed to a higher-order management of interference by integrating information into a coherent situation model of text.;It was first hypothesized that working memory is a major factor impacting the differences in naming patterns on the primary predictive inferences (H1). Two alternative hypotheses parallel what different theories have predicted. The General Capacity Theory (Engle & Kane, 2004) would be supported if the low-span good readers have slower naming times because it would take the low span more time to shift their attention from the secondary inferential concept to the primary concept (H1-a ), Whereas, the Skill Suppression Hypothesis (Gernsbacher & Faust, 1995) would be supported if the low span poor readers have quicker naming times because they would employ the enhancement mechanism (H1-b). The enhancement mechanism refers to when readers are not engaged in deep text processing (i.e., making primary predictive inferences in this study) but accept the superficial meaning of text during reading. Second, it was hypothesized that there would be a significant interaction between the text elaboration and working memory (H2). Third, there should be no significant working memory effect on exact recall of the predictive inference (H3). Finally working memory should contribute to higher-order learning from text ( H4).