Semantics and cognition

Abstract

Introduction This chapter is addressed to semanticists who are interested in studying semantic phenomena using neuro-imaging techniques and who wonder what might be the role of (formal) semantic theorising in such an endeavour. We first look at a domain (discourse) which, due to its temporal component, lends itself particularly well to the technique known as event-related potentials (ERPs). In the second part, we present examples of this technique, and we highlight the importance of formal semantic theories in producing hypotheses testable by ERPs. The third part ends the chapter on a cautionary note and discusses several methodological pitfalls. Discourse Most formal theories of semantics have come around to the view that discourse, not the sentence, is the primary unit of semantic analysis. The reasons are well known: the interpretation of verb tense requires (verbal and non-verbal) context beyond the sentence boundaries, as does the interpretation of anaphora. Kamp's treatment of these phenomena introduced a level of representation mediating between syntax and the first order models standing in for the real world: the discourse representation structure , or discourse model. Although these structures were not deemed to have cognitive significance at the time, contemporaneous developments in psycholinguistics strongly suggested the following hypothesis: language comprehension consists in the construction of a discourse model. It is the purpose of this chapter to show how this view of semantics provides a bridge to cognitive neuroscience: it yields predictions about the shape of neural signals recorded during discourse comprehension, and electrophysiological data to some extent constrain formal theories of discourse representation. Assuming the reader knows these formal theories, we concentrate on cognitive aspects of discourse representation. In understanding discourse, people construct the ‘gist’ of what is said from non-linguistic information available in context, from the language of the discourse, and from prior knowledge and belief. If they cannot construct a gist, they cannot understand the significance of the discourse, or remember it. Gist is a pre-theoretical term for whatever this structure/content is that is derived from a full understanding of a discourse. Several fields have addressed themselves to analysing gist.