Post-nasal devoicing in Tswana

Tswana is traditionally described as having a process of post-nasal stop devoicing (/mba/ → [mpa]). If this description is accurate, then Tswana poses a challenge to views that neutralization processes should be articulatorily grounded. Airflow leakage through the nasal cavity should promote, not inhibit, voicing post-nasally. Zsiga et al. [1] performed an acoustic study of the speech of 6 Tswana speakers, and found no evidence of post-nasal devoicing. They conclude that, counter to the traditional descriptions, Tswana does not have post-nasal devoicing. In an independent study, we collected speech samples from 12 Tswana speakers. Four of our speakers showed clear and consistent post-nasal devoicing. In this paper, we present the data for these 4 speakers to show that at least some speakers of Tswana do have an active process of post-nasal devoicing. We also consider possible explanations for this process, arguing that it is motivated by perceptual rather than articulatory considerations.