A Cognitive Model of Coherence-Driven Story Comprehension

Current models of story comprehension have three major deficiencies: (1) lack of experimental support for the inference processes they involve (e.g. reliance on prediction); (2) indifference to 'kinds' of coherence (e.g. local and global); and (3) inability to find interpretations at variable depths. I propose that comprehension is driven by the need to find a representation that reaches a 'coherence threshold'. Variable inference processes are a reflection of different thresholds, and the skepticism of an individual inference process determines how thresholds are reached. 1 I n t r o d u c t i o n Recent research in psychology maintains that comprehension is 'explanation-driven' (Graesser et al., 1994) and guided by the 'need for coherence' (van den Broek et al., 1995). The comprehender's goal is construction of a more-orless coherent representation which includes explanations for and relations between the story's eventualities. This representation is generated via inferences, which enrich the representation until it reaches the threshold specified by the comprehender's coherence need (van den Broek et al., 1995). By contrast, early models of comprehension emphasised its expectation-driven nature: prediction of future eventualities, followed by substantiation of these predictions (DeJong, 1979). The inference processes described in these early models are still implemented in many contemporary systems. One problem with these models is their failure to account for experimental evidence about inferences: predictive inferences are not generated at point x in the story, unless strongly supported by the story up to point x (Trabasso and Magliano, 1996); in addition, predictive inferences not immediately confirmed by the story after point x are not incorporated into the representation (Murray et al., 1993). While it is difficult to define 'strong support' or 'confirmation', it is clear that an overly-assumptive model does not reflect mundane comprehension. A second problem is the failure of these models to account for differential establishment of local and global coherence. Local coherence holds between 'short sequences of clauses', while global coherence is measured in terms of 'overarching themes' (Graesser et al., 1994). McKoon and Ratcliff (1992) maintain that only local coherence is normally established during comprehension (the minimalist hypothesis). Others state that readers 'at tempt to construct a meaning representation that is coherent at both local and global levels' (the constructionist hypothesis) (Graesser et al., 1994). Script-based models allow globally-coherent structures to be constructed automatically, contradicting the minimalist hypothesis; the inclusion of promiscuous predictive inferences also contradicts the constructionist hypothesis. A third problem is that previous models deny comprehension's flexibility. This issue is sometimes side-stepped by assuming that comprehension concludes with the instantiation of one or more 'primitive' or 'top-level' patterns. Another approach is to apply lower-level patterns which account for smaller subsets of the input, but the aim is still to connect a story's first eventuality to its last (van den Broek et al., 1995). This paper describes a model which treats inferences as coherence generators, where an inference's occurrence depends on its coherence contribution. Unusual inference-making, establishment of local and global coherence, and variable-precision comprehension can be