Understanding text: Achieving explanatory coherence through on‐line inferences and mental operations in working memory

Abstract The question of whether global, causal inferences are made during comprehension is studied following a three‐pronged approach. First, potential inferences were first identified a priori in texts generated by the causal network model for narrative discourse. Second, verbal protocol data in the form of "talking aloud" during the reading of sentences in stories were used to evaluate whether or not people made the predicted inferences. The global, causal inferences occurred where anticipated by the discourse analysis. The talk‐aloud data also revealed that subjects used four main mental operations during comprehension of a text sentence: maintaining, retrieving, elaborating, and explaining. Third, the talk‐aloud data predicted recognition priming of superordinate goal statements, reading times of sentences, coherence ratings of stories, and long‐term retention of stories. The data are discussed with reference to constructivist versus minimalist processing claims, working‐memory models, and what verbal protocols reveal about processing.