Implicit Causality, Negation, and Models of Discourse

Causality plays an important role in giving discourse its characteristic coherence. This paper examines how causality implicit in an utterance helps to organize dynamically constructed mental models of discourse. Experiments are reported suggesting that the linguistic form of utterances contributes significant semantic information about causality to a discourse representation. This view is contrasted with competing claims in the literature that causality only emerges from social psychological inferences or optional inferences on background knowledge. Implicit Causality in Discourse A satisfactory account of the cognitive processes involved in language comprehension must provide an explicit understanding of how individuals make sense of a discourse (i.e., a sequence of spoken or written utterances), and how that sense is related to the world it is about. There is substantial agreement in the psycholinguistic literature that such an account will identify explicit mechanisms that map successive utterances onto a dynamic mental model and research on formal semantics adds to this view the requirement that such a model represent objects in the world and relations between those objects (cf. Johnson-Laird, 1983). One important line of research in this effort attempts to ascertain the form of the mental model of a discourse (i.e., discourse model) and poses questions concerning how it is constructed from linguistic input (e.g., Gordon, Grosz, & Gilliom, 1993; Gordon & Scearce, 1995). A second line of inquiry emphasizes how inferences beyond the information directly represented in the linguistic input play an important role in the elaboration of a discourse model (e.g., Graesser, Singer, & Trabasso, 1994). These two research programs are complementary: the first emphasizes the contribution of the “said” (information taken to be directly represented in the linguistic input); the second emphasizes the contribution of the “unsaid” (information not directly represented in the linguistic input but inferred). In previous work we have examined why natural languages make use of proper names and pronouns to refer to entities in the world (Gordon & Hendrick, 1997). Formal logic constructs formal languages that make no use of pronouns, and our research has been aimed at understanding the role of names and pronouns in dynamic discourse models. Our research suggests that the difference in the linguistic form of referring expressions marks and exploits the orderly elaboration of a discourse model. Common sense suggests that causal connections between utterances of a discourse are also important in establishing a coherent discourse model, and evidence exists that substantiates this point of view (Givón & Gernsbacher, 1995). Subjects can often identify a cause in a single utterance. For example, in the sentence Robin wrote to Sandy, Robin appears to be the cause of the event of writing. This kind of judgment goes under the label implicit causality in the psycholinguistic literature. In this paper we explore the determinants of implicit causality in a discourse model. Three Views of Implicit Causality The psycholinguistic literature distinguishes three conceptions of the source of implicit causality and its integration in a discourse model: 1. Formal semantics has a tradition of categorizing predicates into simplex functions expressing states and complex functions involving a relation of causation and a resulting state (e.g., Dowty, 1975, Parsons, 1990). Some research has built on this semantic tradition and analyzed causality as a relation expressed by a predicate invoked in an utterance of a discourse. To comprehend an utterance is to understand the relation of causality asserted by invoking a particular predicate. Garvey and Caramazza (1974) and Garvey, Caramazza and Yates (1975) offer a typology of predicates based on whether the subject noun phrase or object noun phrase of a particular predicate is treated as the source of implicit causality in an expressed event. 2. Social psychology has led some researchers to analyze the causality implicit in an utterance as originating outside of the linguistic material of a discourse and place it in more general cognitive schemas. Brown and Fish (1983) and Brown (1986) for example observe the same general typology of predicates as Garvey and Caramazza but derive its existence from general, independent principles (of attribution theory) rather than from the semantic nature of linguistic predicates. 3. Non explicit inferences generated from discourse models have also been suggested as likely candidates for judgments of implicit causality. Garnham, Traxler, Oakhill, and Gernsbacher (1996) offer an "integrationist model" in which inferences about implicit causality are derived from discourse models. These inferences are only produced when