Identification of steady‐state sung vowels

Abstract

Sung vowels provide an excellent opportunity to study the information present in the vowel steady state. Ten American English vowels /i,I,e,ɛ,æ,ɑ,ʌ,o,U,u/ in /b/‐vowel‐/d/ syllables were sung by; a professional male singer at seven different pitches: C3, E3, A3, C4, E4, A4, and C4. Syllable length ranged from about 500 to 900 ms. An algorithm was used to extract an approximately 200‐ms sample of the steady‐state portion, carefully avoiding initial and final consonant transitions. Listeners attempted to identify the vowel using only this steady‐state information. Removal of the consonantal context increased vowel identification errors. In addition, identification accuracy was lower for vowels sung at the higher pitches. The error rate was greater for back vowels than for front vowels, especially at the higher pitches. Confusions between /ɑ/ and /ʌ/ were especially common. The effect on vowel identification of using a “chest” voice versus a “head” voice was also investigated. Results are discussed with regard to the adequacy of steady‐state information for vowel identification. [Work supported by NIMH, NICHHD, and NSF.]