Transfer of Statistical Learning from Speech Perception to Speech Production

 The idea that neurocognitive processes subserving speech perception and production are linked  is far from controversial. Yet little is known about the processes comprising these links or how  they operate. One major exception involves talkers’ perception of their own speech. Empirical and  theoretical work continues to provide increasingly sophisticated accounts of the mechanisms  supporting auditory feedback processing of our own voice for the purposes of monitoring and  maintaining ongoing speech production. However, these accounts do not easily explain empirical  findings that other talkers’ speech also affects our own speech productions. My dissertation  introduces a novel experimental approach for studying links between speech perception and  production through the lens of statistical learning across short-term speech regularities that  convey an accent that departs from local speech norms. I begin by demonstrating how  manipulation of short-term statistical regularities conveyed by passive exposure to speech leads  to perceptual statistical learning that drives changes in talkers’ own speech productions. This  transfer of statistical learning from speech perception to production shows how subtle, implicit  alterations in perceived speech can lead to predictable, and robust, changes in participants’ own  speech productions. Following this initial demonstration, I investigate how task demands used to  elicit speech productions affect transfer. Next, I examine whether auditory feedback from one’s  own voice is necessary for transfer from statistical learning. Finally, I test whether generalization  of dimension-based statistical learning in perception across word pairs transfers to influence  speech production. The concluding sections of my dissertation integrate findings across these  investigations to show the promise of statistical learning paradigms in revealing mechanistic links  between speech perception and production.