Files
Abstract
An extensive amount of work has shown the way the working memory architecture supports language comprehension. However, while language comprehension inevitably takes place in a discourse context, less is known about how information related to discourse is organized and managed in working memory. The project particularly focuses on two known processes in the working memory involved in language comprehension, namely encoding and retrieval. I investigate the way discourse structure information guides retrieval and how the information affects the encoding process in real-time comprehension.
To this end, I use two types of sentence structures that are minimally different in their discourse status. While both sentences in (1) contain the information that the waitress is “Kelly’s sister,” the discourse status of such information differs when it is contained in a restrictive relative clause (RRC) (1a) or an appositive relative clause (ARC) (1b). While the content inside an RRC (1a) is part of primary, essential information (i.e., main discourse information), it becomes secondary, and side-commentary (i.e., subordinate discourse information) when it is contained inside an ARC (1b).
(1a) The waitress who is Kelly’s sister sat next to Bob.
(1b) The waitress, who is Kelly’s sister, sat next to Bob.
Given the contrast of main vs. subordinate discourse information status, I examine how this division affects the retrieval process and what effects this information results in during the encoding process. To address these questions, I make use of two of the well-established linguistic phenomena in the sentence processing literature to examine the contrast: number agreement attraction effect, and pronoun resolution. I conduct a series of experiments using behavioral measures such as reading times using a self-paced reading task and eye-gaze times implementing a visual world paradigm.
The findings suggest that discourse structure information imposes a constraint on the memory retrieval process. Specifically, the active state of discourse question in the given discourse structure guides which linguistic entities can be targeted for retrieval. Furthermore, the results show that discourse structure information affects retrieval even when it is not necessarily used for governing the grammaticality of linguistic dependency. While the distinction in discourse status does not affect grammaticality for resolving the dependency, it impacts the ease of dependency resolution in real time. The results also suggest that the parser actively uses discourse structure information during the encoding process, to the extent that the overlap of this information between encoded linguistic representations leads to a competition, encoding interference effect.