Kinematic evidence for the existence of gradient speech errors

Abstract

Speech errors have long been appealed to as evidence for segmental units in speech production. However, if speech errors can be shown to be potentially gradient, as opposed to always categorical, in nature, the view of segments such as ‘‘swapping places’’ must be called into question. Severe limitation of a transcription approach to evaluating speech errors is that gradient errors may go unrepresented in the transcriptional record if they are obscured by other articulatory events due to the coproduction of speech gestures. In a pioneering production study of errors, Mowrey and MacKay [J. Acoust. Soc. Am. 88, 1299–1312 (1990)] present EMG data that suggest that gradient errors do exist, although they are not always audible. However, the anomalous activity they observe in single-motor-unit recordings does not preclude the possibility that a segmental unit was either still produced correctly via compensation by other muscle activity or omitted in its entirety in the articulatory kinematics. Using magnetometry, the present study for the first time is able to provide articulatory movement tracking data that exhibit gradient errors in phrases like ‘‘cop top’’ and ‘‘Bligh Bay’’ repeated rapidly. For example, during the [t] in ‘‘cop top’’ some tokens show a small (sometimes inaudible) raising of the tongue dorsum. [Work supported by NIH.]