In a discourse, not all information is equally important, and elements that are important at one point, may not remain that way as the discourse continues. A story can, for example, focus on a single referent for a while, but then introduce a new referent and zoom in on them – in the meantime, the first referent can remain in the picture or can be dropped altogether. The dynamic nature of discourse requires that language users keep track of referents, which includes keeping track of referents’ accessibility and relative prominence. This information is vital for the production and interpretation of referring expressions: While highly prominent discourse referents can be picked up by reduced referring expressions, less prominent referents can only be referred to using full descriptive terms. It has long been accepted that a referent has to have been at least somewhat recently mentioned in the discourse to be able to be picked up by a reduced referring expression. However, it seems that this effect of recency is mediated by the structure of the discourse.

I will present a series of four pronoun resolution experiments, in which we have examined how the accessibility of referents is impacted by the form of subsequent discourse segments, investigating effects of syntactic subordination, the presence of explicit coherence markers, and typographic and prosodic boundaries. Our findings indicate that syntactic subordination, connectives, and typographic boundaries all additively contribute to whether an intervening clause is perceived as less or more conceptually integrated, and that this affects how strongly that clause blocks access to a preceding referent. However, the type of prosodic boundary was found to interact with syntax in an unforeseen way, which will be elaborated on in the talk.