Joint information value of syntactic and semantic prominence for subsequent pronominal reference

Abstract

Entities realized in prominent syntactic positions receive some preferential treatment when referred to in a subsequent discourse segment: In particular, they are preferentially referred to with reduced referring expressions, often pronouns. This has been observed in both production (Arnold, 1998) and perception (Gordon et al., 1993; Almor, 1999). As a result, syntactic prominence has been regarded as a primary factor in determining the salience of entities. However, in English, syntactic role and semantic role are often conflated. That is, entities realized as syntactic subjects are often semantic agents while those realized as objects are often patients. Thus, it could be that salience is determined by semantic prominence rather than syntactic prominence. In this paper, I present some evidence from a corpus investigation comparing the relative contribution of syntactic and semantic prominence to the salience of entities for subsequent pronominal reference. Using the concept of the value of information from Information Theory (Shannon, 1948), results suggest that syntactic and semantic prominence are comparably informative, but that together they are more informative than either is alone.

References

“…Provided that the meaning of a noun can be coerced in context (Krifka 1995;Talmy 2000;Wisniewski 2009) and syntactic prominence is one of the primary factors in determining the salience of entities (Rose 2011), I assume that gender selection in definite NPs could be driven by the need to mark different degrees of …”