A significant portion of the information conveyed by a typical linguistic utterance is not directly encoded in the words and that comprise it and the way they are articulated, but is rather "added on" via reasoning about how the utterance relates to information that is already present in the context. This feature of language gives rise to ambiguity, vagueness, and other kinds of "interpretive uncertainty", and has often been viewed as a design flaw that must be eradicated from rigorous discourse. In practice, however, humans are exceptionally good at arriving at the right meanings of utterances in the right contexts. The goal of this research project is to investigate exactly how this is accomplished, by examining how the cognitive systems devoted primarily to language interface with more general cognitive systems involved in representing knowledge about the world in the online construction of meaning.
The research conducted by Dr. Kennedy and his colleagues will use experimental techniques to identify and compare the processing signatures of two varieties of context-dependent meaning. The first variety is based primarily in features of the linguistic system, and involves the use of linguistic expressions that are underspecified for content (such as pronouns). The second builds on more general knowledge of the world and the nature of communication, and involves reasoning about speakers' intentions in using particular expressions as opposed to others in a context to achieve a communicative goal. The main experimental method to be employed in the research is known as the Visual World Paradigm, in which experimental subjects' eye-movements are monitored as they follow spoken instructions to manipulate objects in a display. Eye-movements directed at objects in the display are precisely time-locked to information about those objects arriving in the speech signal, and so provide insight on the dynamic construction of meaning down to a granularity of milliseconds. This interdisciplinary research will deepen our understanding of how listeners use both contextual and linguistic clues to build up meanings, and provide resources for researchers in other fields developing tools and techniques for facilitating linguistic communication.
