Dissertations, Theses, and Capstone Projects

Date of Degree

1986

Document Type

Dissertation

Degree Name

Ph.D.

Program

Speech-Language-Hearing Sciences

Advisor

Helen Smith Cairns

Committee Members

Robert W. Fiengo

Michael Studdert-Kennedy

Subject Categories

Speech and Hearing Science

Abstract

Two experiments were conducted to test the SVO Expectancy Hypothesis. This hypothesis embodies three claims: (a) the syntactic parsing device has an initial expectation for the SVO (syntactically defined) structure, (b) the parser reads or tracks the syntactic information in the utterance to confirm or adjust its predictions, and (c) the parser has the ability to make on-line revisions based on the syntactic information contained within the utterance.

In the first experiments using tachistoscopic presentation, 75 sentences representing 15 different sentence types were read by subjects. Each type varied in structure and in the clarity of markers, used to indicate interruptions or deviations from SVO structure. Comprehension time was measured. The results supported the hypothesis that the parser initially expects (predicts) an SVO structure. Furthermore, there was some tentative support for the claim that syntactic information which unambiguously marks the SVO interruptions or deviations facilitates parsing.

The second experiment employed the phoneme monitor task to test both the expectancy and the marker claims on-line. Sentences containing embedded subject and object relative clauses were presented auditorily to subjects. In addition, sentences containing object complements with and without selection restrictions and pragmatic constraints violated were presented to determine the level of processing at which semantic information is used. Experiment 2 failed to yield conclusive results. The failure was attributed to the inappropriate use of a nonmodular task to tap an informationally encapsulated system (parsing). Furthermore, an unanticipated materials effect may have contributed to the failure of the second experiment. Error data and follow-up probe tasks provided support for the SVO effect. In addition, these data suggested that selectional information is not read by the parser but subcategorization information is.

It was concluded that the results of Experiments 1 and 2 support the SVO Expectancy Hypothesis. It was suggested that future research focus on the marking system and the revision process.

Comments

Digital reproduction from the UMI microform.

Share

COinS