| Several predictions are tested regarding a process model of narrative comprehension. The Causal Inference Maker (van den Broek, 1990, in Balota, Flores d'Arcais, & Rayner) describes the logical and procedural conditions under which a reader infers causal relations. As new text is encountered, the reader considers whether those parts of the prior text that are still in the focus of attention explain this new material. When a sufficient explanation is missing, the reader tries to establish information that will provide an explanation and repair the break in coherence. Text material that is no longer in the reader's attention, but can still be retrieved from memory, may offer the needed information. If so, the reader will reinstate this material to establish coherence. If the prior text does not provide the needed information, the reader will use background knowledge to construct an adequate explanation. Previous research supports this model but has not verified that the predicted inferences are made during the reading process.;To test the hypothesis that readers reinstate earlier text information in order to establish a sufficient explanation for a target statement at the time it is read, two versions of several stories were constructed. In one version, reinstatement of a distant cause is required in order to establish a sufficient explanation for the target statement. In the other version, no reinstatement is required to understand the target. When asked to name a probe word taken from the distant cause immediately after reading the target sentence, subjects were faster in the reinstatement condition than in the condition where no reinstatement was required. This facilitation shows that the distant cause was reinstated to the reader's attention when predicted. A subgroup of subjects, those with low reading-span scores, did not show this facilitation thus indicating that limitations on attentional resources may interfere with reinstatement searches.;The findings support the central hypothesis of the Causal Inference Maker model that, during the reading process, readers reinstate earlier text information that is causally related to the current text when that information is needed to maintain causal coherence in the representation of the text. |
