Brain Mechanisms of Syntactic Processing

SAKAI, Kuniyoshi L.
Dept. of Cognitive and Behavioral Science
Graduate School of Arts and Sciences
The University of Tokyo
Dept. of Linguistics, Cognitive Neuroscience of Language Lab.
University of Maryland, College Park

This project focuses on human knowledge of sentence structure, and aims to bridge the gap between our understanding of linguistic structure at the cortical level and at the level of theoretical and computational models. The central issue is how the brain solves the problem of discrete infinity in human language, the means by which humans are able to use a finite store of linguistic knowledge to create an infinite number of sentences. The research project will be carried out by an international team which brings together a group from Japan with expertise in cellular and cortical mechanisms for information processing, and a group from the United States with expertise in dynamic models of linguistic processes. Through a merger of the neuroscientific and linguistic mechanisms identified independently by the two groups, a detailed linking hypothesis emerges for the encoding of linguistic structures in cortical language areas.

The ‘discrete’ part of the solution consists of a mechanism for representing the combinatorial possibilities of a finite set of individual words. This mechanism is hypothesized to be stored in specific temporal regions, and may be related to cellular mechanisms identified by the Japanese group. The ‘infinite’ part of the solution consists of a mechanism for building and maintaining transient representations of combinations of words and phrases. We hypothesize that this mechanism is supported by the subregions of Broca’s area. Additional mechanisms are responsible for effectively combining these two representational devices – these additional processes must search the store of combinatorial possibilities to identify well-formed structural relations, and to diagnose grammatical errors, in order to determine which structures should be represented. The project will extend an explicit computational model of each of these component processes, and will test experimental predictions about which of these processes are active when and where in the construction of sentence structures. The project pursues the hypothesis that languages with radically different word orders differ only in their finite, lexical store of information, such that the mechanisms for analyzing and storing syntactic structures are fixed across languages.

The project focuses on fMRI, MEG, and ERP recordings of speakers of Japanese and English, in conjunction with behavioral experiments and computational modeling. The first set of experiments investigates a variety of different well-formed structural dependencies, and the mechanisms which allow these dependencies to be represented. These include studies of referential dependencies (pronominal coreference, displacement operations) and agreement dependencies (subject-auxiliary agreement, Japanese noun-numeral dependencies). The second set of experiments investigates the grammatical search procedures that diagnose the well- or ill-formedness of sentences, and contrasts the processing of universal and language-particular grammatical constraints. The third set of studies focuses on the degree of uniformity of structure-building mechanisms across languages with radically different word orders.