Perception and acoustics of voiceless initial stops produced by alaryngeal speakers in Taiwan

This study investigates three aspirated voiceless stop consonants and their unaspirated counterparts produced in Taiwan's Southern Min dialect by 10 laryngeal speakers (LA), 10 speakers with pneumatic artificial larynxes (PA), and 10 oesophageal speakers (ES). Listeners' perceptions of the accuracy of the three groups in producing these six stop consonants were also examined. Each of the voiceless stop consonants /p, pʰ, t, tʰ, k, kʰ/, followed by the vowel /a/, was represented by a Southern Min character and appeared three times on a list in random order. Each character had a consonant-vowel syllable structure, and all were read with Tone 1 (high and level). Voice onset time (VOT) was measured for each stop consonant. A separate group of 20 listeners was then asked to identify the consonants produced by the LA, PA, and ES speakers. The VOT values of aspirated voiceless stop consonants from PA and ES were shorter than those from LA. Although PA and ES had aspiration contrast in their VOT values, perceived accuracy was low for PA (/pʰ/: 16.3%; /tʰ/: 13.5%; /kʰ/: 22.8%) and ES (/pʰ/: 33.7%; /tʰ/: 25.9%; /kʰ/: 31.9%). The mean percentages of correctly identified /pʰ, tʰ, kʰ/ produced by PA were at chance level. In terms of VOT values, it is recommended that, when producing aspirated voiceless stop consonants, PA and ES elongate the duration between the release of the oral constriction and the onset of their vibratory sound source. Moreover, it is also recommended that further investigation into other phonetic properties that support stop contrast is needed.