The role of linguistic contrasts in the auditory feedback control of Speech

Abstract

Speakers use auditory feedback to monitor their own speech, ensuring that the intended output matches the observed output. By altering the acoustic feedback signal before it reaches the speaker's ear, we can induce auditory errors: differences between what is expected and what is heard. This dissertation investigates the neural mechanisms responsible for the detection and consequent correction of these auditory errors. Linguistic influences on feedback control were assessed in two experiments employing auditory perturbation. In a behavioral experiment, subjects spoke four-word sentences while the fundamental frequency (FO) of the stressed word was perturbed either upwards or downwards, causing the word to sound more or less stressed. Subjects adapted by altering both the FO and the intensity contrast between stressed and unstressed words, even though intensity remained unperturbed. An integrated model of prosodic control is proposed in which FO and intensity are modulated together to achieve a stress target. In a second experiment, functional magnetic resonance imaging was used to measure neural responses to speech with and without auditory perturbation. Subjects were found to compensate more for formant shifts that resulted in a phonetic category change than for formant shifts that did not, despite the identical magnitudes of the shifts. Furthermore, the extent of neural activation in superior temporal and inferior frontal regions was greater for cross-category than for within-category shifts, evidence that a stronger cortical error signal accompanies a linguistically-relevant acoustic change. Taken together, these results demonstrate that auditory feedback control is sensitive to linguistic contrasts learned through auditory experience.