Anatomy of Unstressed Syllables

Abstract

In the CVC syllable as synthesized on the Pattern Playback the nucleus is generally a three-formant pattern composed of distinct segments; an initial transition, a steady-state part, and a final transition. The formant frequencies of the steady-state part are those characteristic of the vowel as produced in isolation, while the transitions involve shifts in these formant frequencies that are to be related to the neighboring consonants. This neat tripartite division, although not closely approximated by spectrograms of real speech, affects only negligibly the quality of the acoustic signal, while it facilitates the isolation and manipulation of subfeatures of the pattern. For the most part, syllables of the type described have been synthesized as isolated items rather than as constituents of longer sequences. When they are included in longer sequences it becomes necessary. if one wishes to reproduce the normal rhythm of English, to take account of the fact that spoken syllables vary greatly in their durations, in particular, that syllables in unstressed position are frequently of such brief durations that a division into steady-state and transitional parts cannot be meaningfully accomplished. Examination of spectrograms and experiments with synthetic speech both suggest that for many unstressed syllables—namely those containing the schwa vowel—the character of the second formant may be related primarily to the consonants preceding and following the vowel and not so much to a fixed formant frequency of the vowel itself, and that the first formant provides the major cue to the identification of the vowel. (This research was supported by the Veterans Administration under contract V1005 M1253.)