Abstract
This paper reports the results of a study in which variability of formant frequencies for different vowels was examined with regard to several predictions derived from the quantal theory of speech. Two subjects were required to reproduce eight different steady-state synthetic vowels which were presented repeatedly in a randomized order. Spectral analysis was carried out on the vocal responses in order to obtain means and standard deviations of the vowel formant frequencies. In the spirit of the quantal theory, it was predicted that the points vowel. /i/, /a/ and /u/ would show lower standard deviations than the nonpoint vowels because these vowels are assumed to be produced at places in the vocal tract where small perturbations in articulation produce only minimal changes in the resulting formant frequencies. That is, these vowels are assumed to be quantal vowels. The results of this study provided little support for the hypothesis under consideration. A discussion of the outcome of the results as well as some speculation as to its failure to find support for the quantal theory is provided in the report. Several final comments are also offered about computer simulation studies of speech production and the need for additional empirical studies on vowel production with real talkers.