Vowel quality and musical timbre as functions of spectrum envelope and fundamental frequency.

Abstract

Observers gave magnitude estimations of the difference in sound quality between the members of pairs of vowellike synthesized sounds. The observers in one set of experiments were asked to judge differences in vowel quality; those in an analogous set of experiments, differences in musical timbre. The first sound in each pair—the standard—was an imitation of one of six English vowels. The second sound in each pair—the test sound—was derived from the first by modifications of the fundamental frequency, the formant frequencies, or certain combinations of both. When the fundamental of the second sound was raised an octave above the frequency of the first, the estimates of difference—both in vowel quality and in musical timbre—could be reduced to a minimum, if the frequencies of the two lower formants of the sound with the higher fundamental were multiplied by a factor of 1.10. The judgments both of vowel quality and of musical timbre indicated that small changes in the frequencies of the two lower formants made large differences in quality, that small differences in quality could be attributed simply to changes in the fundamental frequency, and that still smaller quality differences could be attributed to changes in the frequencies of the higher formants alone. Shifts in the spectrum envelopes of sounds corresponding to the compact vowels resulted in greater apparent differences in both vowel quality and musical timbre than comparable shifts in the spectrum envelopes of the noncompact sounds. The results suggest that vowel quality and musical timbre are similar functions of their common acoustic correlates, that perceived invariances in the quality of sound cannot be attributed to learning, and they support a modified “fixed pitch” or formant theory of quality.