Otomanguean languages, spoken in Southern Mexico, are well-known for possessing complex lexical tone inventories (DiCanio & Bennett 2019). Contrastive lexical tone is reconstructed within Proto-Otomanguean, which was spoken approximately 4,000-6,000 years ago (Rensch 1976). Even at the sub-family level 3 level tones are reconstructed, e.g. Proto-Mixtecan possessed three level tones (Longacre 1957) and was spoken approximately 2,000 years ago (Josserand 1983). Though a growing body of research has investigated the recent evolution of incipient tones (tonogenesis), such as in Afrikaans (Coetzee et al 2018), Korean (Silva 2006), Kammu (Svantesson & House 2009), Kurtöp (Hyslop 2009), and Tamang (Mazaudon & Michaud 2008), the greater time depth and diversification of tone in Otomanguean largely prevents researchers from investigating its phonetic precursors. Yet despite the focus on incipient tonogenesis in the recent literature, tones in Otomanguean languages continue to diversify and their conditioning environments can be synchronically observed by examining low-level variation in production in closely related language varieties.
In this talk, I examine how processes of speech reduction co-occur with patterns of tonal variation within two distantly-related Mixtecan languages (Yoloxóchitl Mixtec and Itunyoso Triqui). In the first study, we examined the effects of prosodic boundaries on tone production in Yoloxóchitl Mixtec. We observed two of the rising tones (/13, 14/) to be variably leveled (to /2/ and /3/, respectively). Though leveling is pervasive in tone languages (cf. Hyman 2007), for certain speakers this particular process applies in both reduced, durationally short contexts, which tend to disfavor rising tones (Zhang 2004), and long duration contexts, which do not impede contour tone production. Comparative data from a closely-related language (Alcozauca Mixtec) shows a later stage of this leveling process where cognates with the low rising tone /13/ have been leveled to /2/ (Mendoza Ruiz 2016). The Yoloxóchitl pattern closely resembles the process by which reduced speech variants may begin to occur in non-reduced contexts, leading to sound change (cf. Parrell and Narayanan 2018).
In the second study, we examined variation in the production of the 2nd person singular (2S) clitic in a spoken corpus of texts in Itunyoso Triqui. This clitic is variably produced as either /=ɾeʔ1/ or /=ɾ/, the latter missing the syllable's rime. Depending on the stem tone, the 2S clitic either induces (a) no stem tonal change, (b) pre-clitic tonal raising, or (c) low tone spreading (DiCanio 2016). The corpus data show the reduced clitic to be more frequent specifically in contexts where it conditions tonal changes on the preceding stem. This finding suggests that morphologically- conditioned tonal change facilitates segmental reduction. If speakers can predict the clitic from its effect on the stem's tone, reducing its segmental content is less detrimental to interpreting the morphological content of the utterance. In fact, a parallel and more advanced pattern is observed in the closely related Chicahuaxtla Triqui language, where the 2S clitic has been morphologized as /-t/ (Hernández Mendoza 2017).
Together these findings demonstrate how tonal change may be induced by durational contraction/reduction (c.f. Cheng & Xu 2015) and redundancy of morphological cues (multiple exponence). While these triggers are not specific to tonal change, when they occur in languages with elaborated tonal systems as we observe in Otomanguean, they produce tonal variation and change leading to greater tonal diversification in the family.