PROJECT SUMMARY/ABSTRACT
The speech motor system shows a remarkable ability to quickly and efficiently learn movements based on
auditory feedback. One common manifestation of such auditory feedback-based learning is auditory-motor
adaptation, which can be observed in everyday speaking situations that involve changes in speech acoustics,
for which speakers must learn to compensate in order to maintain intelligible speech. Despite its fundamental
nature in maintaining and fine-tuning speech production, however, there exist crucial knowledge gaps in the
understanding of auditory-motor adaptation. In fact, even basic mechanisms in adaptation, such as which
factors drive the learning (i.e., adapting to compensate for auditory perturbations like formant shifts) and
unlearning (i.e., returning to the baseline movements upon the removal of the previously applied auditory
perturbation) processes remain largely unclear. Previous studies have hypothesized that auditory-motor
adaptation may result from minimizing auditory prediction errors–the discrepancies between predicted auditory
consequences of motor commands and actual auditory feedback. To date, however, this fundamental
hypothesis remains surprisingly underinvestigated. Hence, this proposed research combines computational
(Aim 1) and neurophysiological (Aim 2) approaches to directly examine whether auditory prediction errors are
the primary factor driving speech auditory-motor adaptation. As a part of the first specific aim, several new
versions of a computational model of speech motor control, Feedback Awareness Control of Tasks in Speech
(FACTS) will be developed to simulate auditory-motor adaptation. The new versions of FACTS will be tested
and validated in order to examine the mechanistic role of auditory prediction errors in adaptation. Aim 2 will use
magnetoencephalography imaging to examine Speaking-Induced Suppression (SIS)—suppression of auditory
responses to self-produced speech compared to the responses to passively heard speech—which is thought to
represent auditory prediction errors. Here, whole-brain data driven analyses of SIS will be examined during a
series of auditory-motor adaptation tasks. In both aims, auditory-motor adaptation to both spectral
perturbations (i.e., adaptation to perturbed formant frequencies) and temporal perturbations (i.e., adaptation to
lengthened voice onset time) will be examined. Together, this combined computational modeling and
neurophysiological approach will offer key mechanistic insights into the neural basis of auditory-motor
adaptation. The immediate outcome may provide the first direct evidence for the critical role of auditory
prediction errors in auditory-motor adaptation, with profound implications for theoretical and computational
models of speech motor control. The broader impact of the work may establish a solid foundation for novel
treatment strategies to improve speech motor treatment efficacy in clinical populations.