An evaluation of alternative functional models of narrative schemata

This paper evaluates several models of how people encode and retrieve information from short narrative texts. The models differ in their assumptions on four issues: (1) whether or not propositions are encoded with differential probability as a function of their importance in a narrative structure, (2) whether the representation of the text in memory is hierarchical or heterarchical, (3) whether memory retrieval depends primarily on direct access or on a top-down search process, and (4) whether or not lexical information is retained in memory. Specific assumptions are combined to form 12 alternative memory models. These models are evaluated in an experiment in which subjects attempted to learn four narrative texts. Memory for the texts was tested by both recall and recognition either immediately or after a 1-hour delay. Propositional recall, but not recognition, varied as a function of importance. On the recognition test subjects were able to distinguish statements that had occurred in the texts from meaning-preserving paraphrases of these statements. These data suggest that (1) subjects retain lexical information in memory, (2) narrative schemata provide a frame work for encoding all text propositions, (3) these schemata are hierarchically organized, and (4) the schemata are used for top-down retrieval of information at output time.