The inference theory of discourse reference tracking

The multiplicity of possible referential forms has been traditionally viewed as a problem that has to be solved during language production and that may call for special strategies during language comprehension. Much existing research has focused on either why certain forms are used in certain contexts (the “why” question), or on how different forms are processed (the “how” question). In my talk I will describe an approach that addresses both questions together and that views the multiplicity of referential form not as a problem but as the solution language offers to a problem created by the constraints of serial communication and the architecture of the memory system that is used for representing discourse. This view shares its emphasis on the balance between cost and function with pragmatic theories such as Relevance Theory and Accessibility Theory that address the “why” question. However, unlike many of these theories, the view I argue for espouses a clear and independently motivated view of computational cost that affects the integrative stage of referential processing. The present view also shares its appeal to memory mechanisms and processing stages with many theories that address the “how” question. However, in contrast to many of these theories, the present view emphasizes the role of semantic representation in working memory and views differences between word classes such as full names and pronouns as driven by these semantic factors and mirroring semantically driven differences between expressions within the same word class (e.g., definite descriptions of varying levels of specificity). More specifically, I argue that the memory system that is used for representing discourse is prone to interference and that the existence of reduced expressions such as pronouns (and null pronouns in some languages) provides an optimized solution for repeated reference with minimal memory interference. In this view, pronouns are the solution language developed to the problem posed by the need to communicate sequentially using limited size informational units that have to be coherently linked, within a memory system that is prone to interferences. I will present the results from several recent behavioral and neuroimaging studies that support this view and link discourse reference tracking to parietal regions that are also involved in spatial attention.