Knowledge Activation in Story Comprehension

The purpose of this study was to provide behavioral data regarding the world knowledge that readers activate during story comprehension with a view to inform the development of formal and computational representation of knowledge for automated comprehension. The focus of the study was on the amount, the structural aspects, and the coherence function of the activated knowledge. In a knowledge elicitation task, undergraduate students recorded the knowledge they deemed necessary for understanding each sentence in a set of four short stories in the form of property and causal knowledge rules. Results indicated the activation of a limited amount of knowledge that maintained both situational and conceptual connections with input. Situational inferencing and continuity (i.e., carrying over of previously activated ideas) were positively associated and varied as a function of story part. Findings are discussed in relation to dominant theoretical models of narrative comprehension and for their implications regarding the computational representation of knowledge in automated story comprehension.