The phonetic nature of “voiceless” nasals in Pokomo: Implications for sound change

Several of the diachronic seriations proposed for Bantu languages assume a relatively early process of nasal devoicing which preceded, and set the stage for, development of stop aspiration (e.g., Kerremans 1980; Hinnebusch, Nurse and Mould 1981). Maddieson (1991) has argued convincingly that one synchronic reflex, an aspirated nasal in Sukuma, may have developed without an intervening stage involving a voiceless nasal. However, reflexes reportedly involving voiceless nasals or nasal loss might still be thought to motivate a sedation that posits a stage with voiceless nasals. We will argue that even these reflexes need not have developed from an intervening stage in which the nasal was phonologically voiceless. For one thing, developments in feature theory argue against representations which posit a phonologically voiceless nasal and following aspirated stop. Furthermore, our phonetic data on Pokomo suggest that nasal voicelessness can result from a phonetic effect temporal overlap between the nasal and the aspiration gesture for the following stop. We propose that changes in phonetic detail such as the magnitude and timing of gestures produced nasal voicelessness or partial voicelessness, phonetic conditions which led to phonological restructuring in many of the Bantu languages.