Extreme stop allophony in Mixtec spontaneous speech: data, word prosody, and modelling

Area of Science:

• Phonetics and Phonology
• Linguistic Anthropology
• Computational Linguistics

Background:

• Word-level prosody influences consonant lenition, typically strengthening initial consonants and weakening medial ones.
• Yoloxóchitl Mixtec, an endangered Mexican language, displays significant lenition in voiceless consonants, even in careful speech.

Purpose of the Study:

• To investigate the relationship between word-prosodic position and obstruent lenition in Yoloxóchitl Mixtec.
• To analyze consonant duration, voicing, and manner of articulation in relation to prosodic context.
• To develop computational models for predicting lenition patterns.

Main Methods:

• Analysis of a spontaneous speech corpus of Yoloxóchitl Mixtec.
• Examination of consonant duration, passive voicing, and manner of articulation.
• Application of deep neural networks for pattern modeling and categorization.

Main Results:

• Word-medial pre-tonic position is identified as the primary site for consonant lengthening and reduced voicing.
• Non-pre-tonic consonants exhibit greater voicing and shorter duration.
• Morpheme functional status influences voicing lenition; word-medial pre-tonic stops show more complete closure than non-pre-tonic stops.

Conclusions:

• The study confirms specific prosodic positions driving consonant lenition in Yoloxóchitl Mixtec.
• Findings align with broader research on phonetics of lenition and offer computational tools for analysis.
• The research contributes to understanding endangered languages and developing predictive models for phonological processes.