The Form of Referential Expressions in Discourse

Most instances of real-life language use involve discourses in which several sentences or utterances are coherently linked through the use of repeated references. Repeated reference can take many forms, and the choice of referential form has been the focus of much research in several related fields. In this article we distinguish between three main approaches: one that addresses the ‘why’ question – why are certain forms used in certain contexts; one that addresses the ‘how’ question – how are different forms processed; and one that aims to answer both questions by seriously considering both the discourse function of referential expressions, and the cognitive mechanisms that underlie their processing cost. We argue that only the latter approach is capable of providing a complete view of referential processing, and that in so doing it may also answer a more profound ‘why’ question – why does language offer multiple referential forms. Coherent discourse typically involves repeated references to previously mentioned referents, and these references can be made with different forms. For example, a person mentioned in discourse can be referred to by a proper name (e.g., Bill), a definite description (e.g., the waiter), or a pronoun (e.g., he). When repeated reference is made to a referent that was mentioned in the same sentence, the choice and processing of referential form may be governed by syntactic constraints such as binding principles (Chomsky 1981). However, in many cases of repeated reference to a referent that was mentioned in the same sentence, and in all cases of repeated reference across sentences, the choice and processing of referential form reflects regular patterns and preferences rather than strong syntactic constraints. The present article focuses on the factors that underlie these patterns. Considerable research in several disciplines has aimed to explain how speakers and writers choose which form they should use to refer to objects and events in discourse, and how listeners and readers process different referential forms (e.g., Chafe 1976; Clark & Wilkes 1986; Kintsch 1988; Gernsbacher 1989; Ariel 1990; Gordon, Grosz & Gilliom 1993; Gundel, Hedberg & Zacharski 1993; Garrod & Sanford 1994; Gordon & Hendrick 1998; Almor 1999; Cowles & Garnham 2005). One of the central observations in this research is that there exists an inverse relation between the specificity of the referential