Realizations of Malagasy vowel devoicing

Abstract

This paper investigates the acoustic realizations of so-called “devoiced” vowels in Merina Malagasy. Malagasy has five monophthongs (/aeiou/; Howe, 2019) of which /a/, /i/, and /u/ have been said to be devoiced. This paper represents the first thorough description of the acoustics of these vowels. In a production experiment, speakers pronounced 115 tokens involving /a/, /i/, and /u/ in prosodic environments described as causing devoicing. Preliminary results indicate that so-called devoiced vowels may be realized as one of at least three variants: devoiced, co-articulated, or deleted. When devoiced, which typically occurs following a voiceless obstruent, the vowel is realized as a lengthening of the aperiodic noise associated with the preceding obstruent. When co-articulated, which is common following some sonorants, the vowel is realized as a gesture on the preceding sonorant without taking up its own time slot; for example, /u/ is realized as a lowered F2 on the preceding sonorant, indicating rounding. Finally, a vowel may be fully deleted, in which case the vowel has no acoustic representation, neither as a gestural overlap with, nor lengthening of the preceding consonant.