A multi-method approach to correlate identification in acoustic data: The case of Media Lengua

This study of Media Lengua examines production differences between mid and high vowels in order to identify the major correlates that distinguish these vowel types. The Media Lengua vowel system is unusual in that it incorporates lexical items originating in Spanish’s five-vowel system into a three-vowel system inherited from Quichua, resulting in high degrees of overlap between the front versus back, mid and high vowel pairs /e, i/ and /o, u/ in F1xF2 space. As Media Lengua speakers utilize and differentiate between all five vowels despite the large degree of acoustic overlap between mid and high vowels, this raises the question of what other correlates beyond F1 and F2 might be involved. To address this, our study looks at a range of variables, both acoustic and qualitative, in a multi-method approach using both factor analysis for mixed data and linear mixed effects regression modelling. Each method provides a unique view on the correlates of vowel differentiation in Media Lengua. Taken together, our results indicate that Media Lengua speakers rely on both social and linguistic contextual cues to distinguish mid from high vowels, which overlap in acoustic space (F1 and F2).