Word-medial strengthening in Mixtecan languages
Within the literature on consonant lenition one observes a common phenomenon of consonantal weakening in word-medial position, while consonants produced in word-initial position are either maintained or strengthened (Fougeron & Keating 1997, Gurevich 2011, Katz & Fricke 2018, Keating et al. 2004, Lavoie 2001). Initial strengthening is realized via either durational lengthening or, in the case of stop consonants, greater contact between active and passive articulators. Lenition in word-medial position is typically realized via a pattern of voicing bleed (in voiceless obstruents) or a more general loss of complete contact between articulators for stop consonants (Hualde et al 2011, Katz & Fricke 2018, Katz & Pitzanti 2019, Torreira & Ernestus 2011). A functional theory for this phenomenon is that initial strengthening aids the listener in lexical access and parsing (Katz & Fricke 2018) and could therefore be a linguistic universal (White et al. 2020).
In this talk, I discuss an intriguing pattern of prosodically-conditioned consonant lengthening which occurs in word-medial position in Yoloxóchitl Mixtec and Itunyoso Triqui - both indigenous languages spoken in Southern Mexico. Both languages are heavily tonal with stem-final stressed syllables. Studies on each language involved corpus phonetic methods applied to running speech from language documentation corpora. For Yoloxóchitl Mixtec, obstruents vary substantially in their realization. Here, we find a robust pattern of root-medial consonant lengthening (DiCanio et al. submitted) which replicates a pattern observed in previous studies (DiCanio et al. 2018, 2020, 2021). Stops in word-medial position are more often voiceless and almost always realized with complete closure. By contrast, stops in word-initial position are more likely to be partially or fully voiced and are often realized without complete closure. For Itunyoso Triqui, consonant lengthening is more variable in root-medial position in running speech than in controlled speech (cf. DiCanio 2010) but no pattern of word-initial lengthening is found.
Each language appears to contradict the functional theory for consonant lenition stated above, but instead show a more robust pattern of stress-conditioned strengthening (cf. Bouavichith & Davidson 2013). Though, an important morphological detail also seems to be relevant - words in Mixtecan languages are only prefixal (suffixes do not occur). Lexical access for languages with prefixal structure is generally delayed relative to languages where roots appear earlier in the word (Ussishkin et al 2017). As a result, initial strengthening might not improve lexical access if it is mainly applied to prefixes or to unstressed syllables in languages with fixed final stress, like Yoloxóchitl Mixtec and Itunyoso Triqui. Past work on initial strengthening has been mostly limited to suffixing languages with non-final stress patterns. The results here both suggest that (a) morphological structure and stress position may play a role in determining variable patterns of consonant lenition in human language and (b) word-initial strengthening is not a universal feature of human language.