Voice onset time versus articulatory modeling for stop consonants

Abstract

Voice onset time (VOT) was developed as a parameter for the pattern-playback speech synthesizer developed approximately 50 years ago, in order to generate the acoustic effects of voiced-unvoiced differences in English stop consonants. However, problems arose when the VOT parameter was used to define stops in actual spoken language, to replace aerodynamic and physiological parameters. A representative physiological model from the same time-period that avoided these problems is sketched. In this model, the manner-of-articulation of a stop is determined by the duration, timing, and extent of laryngeal, articulatory, and respiratory gestures. It is concluded that the term voice onset time should be used only as a parameter in speech synthesis, as originally intended, and not for the analysis of actual speech.