Articulation without acoustics: “Soundless” vowels in Oneida and Blackfoot

Speech is largely driven by a feed-forward production mechanism, allowing articulators to reach their targets even in the absence of auditory feedback. However, it is not known whether natural languages can systematically encode such movements without acoustic consequences. Perception and production studies were conducted of two endangered languages purported to exhibit this property in utterance-final sequences. Ultrasound, video, acoustic, and perception data were collected from native speakers of Oneida (Iroquoian) and Blackfoot (Algonquian) in field settings. Results show distinct articulator positions for different utterance-final vowels despite these vowels being completely inaudible, both in terms of acoustic and perceptual measures. Speech production models must include a sufficiently robust feed-forward articulatory component to allow articulations to be realized even when acoustic information is systematically absent.

► Articulatory, acoustic, and perception data collected from Oneida and Blackfoot speakers.
► Results show distinct articulator positions for different inaudible vowels.
► Articulations can be phonologically stable when production lacks acoustic consequence.
► Feed-forward articulatory component would allow articulations with no acoustic output.