The Interface of Syntax, Semantics and Psycholinguistics: An Investigation of Operator Scope

The goal of this paper is to explore an on-line processing theory of discourse comprehension which makes use of the grammatical construct of modality. Previous investigations of discourse processing have relied upon studies using the integrative device of anaphora (Halliday and Hasan, 1976). A close examination of the kinds of discourses previously investigated reveals that the sorts of inferences required to discern the antecedent to a pronoun rely on pragmatic world knowledge only, to the exclusion of grammatical knowledge (cf., Leonard, Waters & Caplan, 1997; Hirst & Brill, 1980; among others). For example, Garrod and Sanford (1994) indicated that either Bill or his friend in the context sentence below can serve as an antecedent to he in a continuation sentence. E.g., Bill1 wanted to lend his friend2 some money. He2 was hard up and really needed it. Another possible anaphoric relation would be: However, he1 was hard up and couldn’t afford to. Assignment of co-reference in these sentences is dependent upon the sentential context. This contextual information refers to what we know about when it is appropriate to need or lend money, and who can be a possible lender. In other words, this is knowledge that is acquired with cultural experience. The question we pose here is whether we can see empirical effects of constructs that come from a different source: the grammar. A discourse processing model that also takes grammatical constructs into account has better predictive power than one that relies exclusively on experiential knowledge, due to potential cross-linguistic predictions. That is, a model which relies solely on world knowledge is bound to fail cross-linguistically, since crosscultural norms can vary tremendously. In the present study, we explored how modality constrains discourse anaphora. Roberts (1987, 1989) developed a theory of Modal Subordination, in the framework of Discourse Representation Theory (Kamp,1981; Heim,1982). Before we explain the phenomenon of how modality constrains anaphoric possibilities, let us first define mood. The mood of an utterance is an indication about the speaker’s commitment to the truth of a proposition in the actual world—that is, it tells us whether or not a proposition is asserted. Factual mood indicates that the Common Ground must be updated, since a true proposition has been added to the discourse context set (Stalnaker, 1978). On the other hand, nonfactual mood indicates that a proposition is introduced as a hypothetical notion, and subsequent propositions might rely on the assumption that the discourse continues as if that previous proposition were true. Factual mood is marked by indicative grammatical mood, whereas non-factual mood has many possible lexical sources: modal auxiliaries (e.g., would, should, can, may, etc.), modal adverbs (perhaps, possibly, maybe, etc.), non-factive propositional attitude verbs (e.g., wonder, consider, muse, etc.). We call these lexical items the class of modal operators (Heim, 1982; Asher, 1987). According to Roberts, these operators define scopal domains, which are represented in the Discourse Representation Structure as subordinate boxes. In contrast, the descriptive content of propositions uttered in factual mood is always entered at the matrix level of the DRS. Thus, factual mood does not add structure to a DRS, unlike non-factual mood. The scopal domains of modal operators thus segment a discourse structure. Roberts (1987, 1989) uses this theoretical construct to account for an observation, originally noted in Karttunen (1976), that indefinite NPs do not form felicitous antecedents to pronouns when they are contained in sentences in non-factual mood but their pronouns are in sentences that mark factual mood. E.g.,