How a Full Account of Segmental Perception Depends on Prosody and Vice Versa

Abstract

The synthesis time-base of a synthetic precursor phrase is one factor determining the position of the phoneme boundary on a continuum of synthetic CV target syllables, which vary in the durational cue Voice Onset Times (VOT) and are introduced by the precursor. Reducing the time-base, thereby increasing the rate of speech in the precursor, increases the probability of the consonant in the target being perceived as voiceless. The results of such perceptual experiments are compared to those of a production study in which six adult male speakers of British English produced examples of CV syllables composed of the consonants /b,p,g,k/ and the vowels /i,a/ in a sentence frame. VOTs in productions of /p/ and /k/ were 20 milliseconds shorter at fast than at slow rates of speech, warranting a normalisation of VOT duration in perception like that actually obtained.

VOTs were longer in /ki/ than in /ka/. It has been suggested that voicing onset is retarded before high vowels because the oral constriction is reduced more slowly than before lower vowels, thus delaying the attainment of the transglottal pressure drop required for voicing. However, VOTs in /pi/ were shorter than those in /pa/. In the case of voiceless bilabials the aerodynamic factor appears to be outweighed by a mechanical influence of anticipatory co-articulation of tongue position which leads to larynx elevation and less vocal cord abduction before high vowels. The production result for velars, but not that for bilabials, is paralleled in perception where longer values of VOT are required for stops to be perceived as voiceless before /i/ compared to /a/ at all places of production. This failure of the perceptual process to follow exactly the constraints in production provides an illustration of a heuristic rather than an algorithmic perceptual strategy presumably designed to allow fast decisions while tolerating some loss of accuracy in exceptional cases.

The ‘precursor target’ paradigm used in these experiments could be extended to examine prosodie influences on other segmental distinctions, but also, to determine the perceptual substrates of prosodie variables such as “rate of speech” and to measure the precision of perceptual expectations for vowel and consonant durations in different sentential and syntactic environments.