Inferential processing during discourse comprehension



Journal Title

Journal ISSN

Volume Title



This study compares two types of models that have been proposed to explain inferential processes employed during the comprehension of written discourse. One of the models, proposed by Thorndyke (1976), suggests prospective and retrospective components of inferential production. The prospective processes infer new propositions that anticipate resultant states. The retrospective processes selectively reinforce or reduce the salience of prospective inferences. The other model alternatively suggests that prospective inferences are not normally made during comprehension. A recognition test procedure similar to Thorndyke's was used to test the two models. Subjects reading passages containing target sentences that suggested prospective inferences were no more likely to falsely recognize these test inferences than were subjects reading passages without the targets. Alternatively, subjects reading passages containing continuation sentences that required retrospective inferences for comprehension significantly increased false recognitions to these test inferences. In contrast to Thorndyke's proposal, these results suggest that people may not normally make prospective inferences that are included in a text representation. In agreement with Thorndyke, these results suggest that people make retrospective inferences and include them in a text representation along with the explicit information contained in discourse.