Durational Characteristics of Word-Initial Consonant Clusters in English

Abstract

A list of 224 English words beginning with prestressed consonant clusters or single consonants were recorded in the frame “Say … instead” by three talkers. Consonantal durations were measured from spectrographic analysis of the data. The results indicate that consonants in clusters are generally shorter than in a singleton environment, but that the amount of shortening is dependent on certain features of adjacent consonants. A system of durational adjustment rules is proposed in order to account for the statistically significant durational differences. If the consonantal duration in the singleton environment is considered as the standard duration, then rules (1) shorten two-element clusters by about 12%; (2) shorten three-element clusters by about 18%, (3) produce additional shortening in a consonant that is followed by a balistic closure; (4) produce additional shortening in a sequence of two voiceless consonants; (5) prevent labials from being shortened as much and shorten adjacent segments somewhat more to compensate; (6) lengthen a consonant if a preceding plosive causes aspiration to be present during a portion of the consonant; and (7) produces additional shortening in a sequence of two dentals. Implications for articulatory modeling, speech perception, and speech synthesis by rule are discussed. [Work supported in part by National Inst. of Health and by Office of Naval Research.]