Date of Award

Fall 2010

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Graduate Group

Linguistics

First Advisor

Jiahong Yuan

Second Advisor

Mark Liberman

Third Advisor

Daniel Swingley

Abstract

The [voice] distinction between homorganic stops and fricatives is made by a number of acoustic correlates including voicing, segment duration, and preceding vowel duration. The present work looks at [voice] from a number of multidimensional perspectives.

This dissertation's focus is a corpus study of the phonetic realization of [voice] in two English-learning infants aged 1;1--3;5. While preceding vowel duration has been studied before in infants, the other correlates of post-vocalic voicing investigated here --- preceding F1, consonant duration, and closure voicing intensity --- had not been measured before in infant speech. The study makes empirical contributions regarding the development of the production of [voice] in infants, not just from a surface-level perspective but also with implications for the phonetics-phonology interface in the adult and developing linguistic systems. Additionally, several methodological contributions will be made in the use of large sized corpora and data modeling techniques.

The study revealed that even in infants, F1 at the midpoint of a vowel preceding a voiced consonant was lower by roughly 50 Hz compared to a vowel before a voiceless consonant, which is in line with the effect found in adults. But while the effect has been considered most likely to be a physiological and nonlinguistic phenomenon in adults, it actually appeared to be correlated in the wrong direction with other aspects of [voice] here, casting doubt on a physiological explanation. Some of the consonant pairs had statistically significant differences in duration and closure voicing. Additionally, a preceding vowel duration difference was found and as well a preliminary indication of a developmental trend that suggests the preceding vowel duration difference is being learned.

The phonetics of adult speech is also considered. Results are presented from a dialectal corpus study of North American English and a lab speech experiment which clarifies the relationship between preceding vowel duration and flapping and the relationship between [voice] and F1 in preceding vowels. Fluent adult speech is also described and machine learning algorithms are applied to learning the [voice] distinction using multidimensional acoustic input plus some lexical knowledge.

Comments

Please see my home page, http://razor.occams.info, for the data files and scripts that make this reproducible research.

Share

COinS