| An enduring debate in comprehension centers around the degree to which readers utilize top-down (knowledge-driven) processes during reading. A test case in this debate is the extent to which readers make predictive elaborative inferences (i.e., those that forecast the outcomes of text-based events). Two theories of inference, minimalism and constructionism, agree that predictive inferences will be made if text is high in causal constraint (i.e., predictability). However, previous research on causal constraint has not distinguished between activating an inference concept and incorporating it into a mental representation.;This research examined whether the degree of causal constraint in narrative text affects the activation and incorporation of predictive inferences. In Experiment 1, a constrained word stem task was used to assess inference activation. The data indicated that readers activate predictive inferences when reading text at high levels of causal constraint, but not at medium or low levels of causal constraint. In Experiment 2, reading times for target sentences that contradicted an invited inference were used to assess inference incorporation. The data suggested that even under conditions of high causal constraint, activated predictions are not incorporated into a reader's mental model of the text. In the final experiment, reading times for a target sentence that confirmed an invited inference were facilitated when readers read text high in causal constraint, suggesting that activated inferences that fail to be incorporated may still exert top-down effects on comprehension. The results favor a cognitive control view of the flexibility of reading processes in comprehension. |
