Causal thinking and the representation of narrative events

Abstract

The basis for representing narrative events in memory was investigated in reanalyses of the stories and data of R. C. Omanson (1982b, Journal of Verbal Learning and Verbal Behavior, 21, 326–337) and N. L. Stein and C. G. Glenn (1979, In New Directions in Discourse Processing, Hillsdale, NJ, Erlbaum). Causal network representations of the stories were derived for prediction of data on immediate and delayed recall, summarization, and judged importance of events. Properties of the networks were compared in multiple regression analyses with other factors, notably the story-grammar categories of the events. Whether or not an event was in a causal chain and the number of its causal connections were both found to account for substantial proportions of common and unique variance in all four measures. The story-grammar category of events also contributed unique variance but overlapped substantially with the causal factors. The concreteness, serial position, and argument overlap of an event failed to account uniquely for the data. A recursive transition network model is discussed that integrates story grammar, causal chain, causal network, and hierarchical problem-solving approaches to story representation.