The role of predictive inferences in situation model construction

Abstract

Current theories of discourse comprehension assert that understanding entails the construction of a multilevel representation of text, which would include a surface structure code, a propositional textbase, and a situation model. The component processes used to generate the surface code and the propositional representation have been specified, yet the processes used in the generation of the situation model have not been identified. These experiments investigated the hypothesis that one of the processes used in situation model development is an inferential process in which the reader strategically employs knowledge to form expectations when highly predictive text opportunities arise. Data from a word recognition task demonstrated that predictions are not encoded into a reader's propositional representation during reading. However, the results from word prediction and lexical decision experiments suggested that predictions are available to the reader at a more abstract level of representation—the situation model.