Causal elaborative inferences in text comprehension: Implications for psychology and education

This thesis explores the role of elaborative inferences in learning, and focuses on the special role of causal structure in improving learners' memory for text. It investigates whether amount of causal sufficiency in narrative text affects the likelihood that readers generate specific target predictions, called forward inferences. It also asks what happens when readers do not generate the predicted target, and if reading ability determines whether a reader automatically makes such inferences. Three experiments examined whether causal theory predicted forward inferencing in both strategic and automatic testing conditions. All experiments used the same items. In the first experiment, subjects were given one-sentence scenarios that varied in degree of causal sufficiency, and asked to write what they thought would happen next. Subjects' sensitivity to degree of Sufficiency was supported in two ways. First, the greater the degree of sufficiency, the more likely subjects were to generate the correct target inference. Second, when subjects did not generate the target, they wrote answers that implied the target's occurrence. The second experiment determined that any differences by degree of sufficiency were not due solely to word-activation, but were generated from the situation described by the scenario as a whole. The third experiment sought converging evidence with an on-line lexical decision task; subjects' reading ability was also measured with Daneman and Carpenter's (1980) reading span test. There were no differences in reaction time, nor accuracy by sufficiency, nor by reading span. Thus, causal structure determines whether readers make specific forward inferences. In addition, these forward inferences are in the form of general hypotheses about what is occurring in the situation described, as well as specific predictions of what will happen next. This does not happen immediately, however. In addition, it is argued that the definition of causality used in these studies yields a more fine-grained analysis of readers' elaborations, not possible with the all-or-nothing definitions characteristic of other text comprehension research. Further studies are proposed to investigate how long it takes for subjects to automatically generate these inferences, as well as what other textual and learning conditions foster forward inferences, in both psychology and education.