STUDIES OF HUMAN SYNTACTIC PROCESSING: RANKED-PARALLEL VERSUS SERIAL MODELS

Date of Completion

January 1987

Keywords

Language, Linguistics

Degree

Ph.D.

Abstract

A central characteristic of many current models of human sentence processing is the manner in which the parser responds to structural ambiguity. Let us define an input string as structurally ambiguous if it is compatible with more than one syntactic structure. Then, following Fodor, Bever and Garrett (1974), we can distinguish two general processing models: (i) parallel models which construct multiple analyses corresponding to the various readings of the ambiguity, and (ii) serial models which construct a single analysis and must reanalyze if this structure is incompatible with subsequent lexical material.^ Despite the great deal of attention which this issue has received in the psycholinguistic literature, there is still no general agreement among researchers concerning how the parser responds to structural ambiguity. Previous research has failed to provide a clear account of ambiguity resolution, with some studies appearing to yield results supporting serial models, and others appearing to support the parallel hypothesis.^ Using a syntactic priming paradigm, this dissertation presents new experimental evidence which indicates that the parser constructs, in parallel, multiple syntactic representations for structurally ambiguous input strings. It also outlines a model of human sentence processing which seeks to reconcile the present results with the findings of recent studies which have been interpreted as support for serial processing models. Additional experiments compare the syntactic priming paradigm with the acceptability judgment task used in many previous studies. Although the proposed model of natural language processing is, at present, underdetermined by the available experimental data, it can serve as a framework for future research into basic issues involving the parser's response to structural ambiguity. ^

Share

COinS