Prosody and speech perception

Abstract

The major concern of this thesis is with models of speech perception. Following Gibson's (1966) work on visual perception, it seeks to establish whether there are sources of information in the speech signal which can be responded to directly and which specify the units of information of speech. The treatment of intonation follows that of Halliday (1967) and rhythm that of Abercrombie (1967) . By "prosody" is taken to mean both the intonational and the rhythmic aspects of speech.

Experiments one to four show the interdependence of prosody and grammar in the perception of speech, although they leave open the question of which sort of information is responded to first. Experiments five and six, employing a short-term memory paradigm and Morton's (1970) "suffix effect" explanation, demonstrate that prosody could well be responded to before grammar. Since the previous experiments suggested a close connection between the two, these results suggest that information about grammatical structures may well be given directly by prosody. In qthe final two experiments the amount of prosodic information in fluent speech that can be perceived independently of grammar and meaning is investigated. Although tone -group division seems to be given clearly enough by acoustic cues, there are problems of interpretation with the data on syllable stress assignments.

In the concluding chapter, a three-stage model of speech perception is proposed, following never (1970), but incorporating prosodic analysis as an integral part of the processing. The obtained experimental results are integrated within this model.