Prosodic signals of sentence structure: Their syntactic distribution

Abstract

We have previously reported a technique for the use of perception tests to identify boundary signals in continuous reading. The distribution of boundary locations between words was described generally in terms of simple parsing state changes, or lack of change. Now we are beginning to analyze the classification of boundary placement in considerable syntactic detail. Computer storage of the data has made it possible to search quickly for various sets of information about syntax and perception. The present data base for this continuing study includes the responses of three listeners to three speakers, each reading the same 1870-word text. Because of the sequential nature of the speech stream in time, boundary signals can be imposed on the words and sentence segments only successively, as they occur. But additional analysis suggests the way in which this sequential system realizes the multidimensional combination of sentence segments. At each boundary, there exists a limited range of possibilities for the succeeding segment, with different probabilities of occurrence or different degrees of functional relatedness attached to each. The least familiar will be most highly marked. The set of possibilities at each kind of boundary may be stated in syntactic terms.