Armenian schwa: a phonetic and phonological analysis

Eastern Armenian (EA) has been described as having stress on the final full vowel of a word (i.e., non-schwa) (Dum-Tragut 2009, Hulst 1999, Khachatryan 1988, Vaux 1998), making the language a sonority-driven stress system (Kenstowicz 1994, de Lacy 2004). The phonological evidence involves high vowel reduction in derived environments (Vaux 1998), while the phonetic evidence demonstrates lengthened vowel duration consistent with the sonority-driven stress pattern (Khachatryan 1988). The current study reanalyzes the phonological evidence as motivated independent of stress, and provides evidence from an acoustic analysis that, on the contrary, EA is more accurately analyzed as having a consistently final metrical head. The study involves a production experiment whose subject is a native EA speaker who read aloud a number of disyllabic CVCVC words in two frame sentences which differed in whether the target was focused. Duration (normalized for speech rate), intensity, F0, and quality (F1/F2) were measured for each target vowel using Praat (Boersma & Weenink 2015) and VoiceSauce (Shue et al 2011). Results were evaluated using standard statistical models (R Core Team 2015). The finding is that there is no distinction in duration, intensity, and vowel quality between putatively stressed and unstressed vowels. There is, however, a pitch rise over the final syllable. This pitch excursion is analyzed as involving an LH* tonal contour, where the high pitch accent associates with a word-final metrical head. The word-final LH* contour occurs consistently, regardless of vowel quality. Ultimately, the study casts doubt on impressionistic reports of stress, which are possibly tonal or pitch accent systems (see e.g., Gordon 2014).M.A.Includes bibliographical referencesby Vartan Haghverd