The implementation of voicing in obstruents in American English connected speech

Abstract

Descriptions of obstruent voicing in English phonetics textbooks observe that [ + voice] obstruents are likely only fully voiced between other voiced sounds, but have less voicing in word-initial and word-final position [e.g., Cruttenden (2008); Docherty (1992) for experimental data for British English]. This study examines these claims for American English using a corpus of 27 speakers reading 3–5 short stories. Voiced stops and fricatives were coded for preceding and following segments, stress of the previous and following syllables, position within the word (initial, medial, and final), and phrasal position of the word. The calculation of unvoiced frames from the Praat voice report was used to obtain the proportion of voicing during each obstruent closure. Results indicate the influence of several phonotactic and prosodic factors. (1) The segment preceding the obstruent influences the amount of voicing more than the following segment does; complete devoicing and shorter durations of partial voicing are most often conditioned by preceding pauses or voiceless sounds. (2) Among following sounds, nasals condition the most complete voicing. (3) An obstruent's position within a word has less effect than its phrasal position. (3) Stress primarily affects stops: a preceding stressed syllable conditions significantly more voicing than a preceding unstressed syllable.