Doctoral Dissertation Research: Investigating generalization, transfer, and representation resulting from non-native speech category training

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Globalization has increased the importance of information transfer across traditional language boundaries. This has created a demand for new approaches to adult language learning that are both efficient and effective. This need is real; there are more second-language speakers of English in the world than native speakers. Bilingualism also confers a competitive advantage with respect to enhanced cognitive abilities and protection from neurodegenerative decline, so developing next-generation approaches to promoting bilingualism may have even broader impact. One of the significant challenges to this endeavor is teaching adults the sounds of a new language. The speech sounds that form the building blocks of language vary considerably across languages. A long history of laboratory research demonstrates that the brain "commits" to the speech sounds of the native language. This supports efficient native speech processing but confers a cost to learning a new language that uses a different sound system. Perceiving non-native speech sounds in adulthood is a case where it has been thought that the adult brain simply cannot be trained to learn. This consensus has arisen from decades of attempts to devise effective learning regimens for training adults to improve non-native speech perception. To summarize this literature very briefly, learning gains are modest even with extensive training over the course of weeks. Moreover, learning does not generalize well to unfamiliar instances of the sounds or to new voices. Under the direction of Dr. Lori Holt, Mr. Liu will target a promising new approach to this problem. Prior research from this team led to the counterintuitive finding that directing attention away from the task of language learning can lead to more efficient learning gains than the traditional approach of making overt judgments and getting explicit feedback. Examining adult learning when training is "hidden" in other engaging tasks is a novel and potentially more effective approach to teaching adults new language sounds. Mr. Liu's dissertation project builds on proof-of-concept evidence within this implicit training task. This study will examine whether extended (1-week) implicit perceptual training targeting improvement of speech sound representations impacts underlying neural representations (assessed via electrophysiological brain responses) and transfer to improvements in vocabulary-learning, modeled after classroom learning. Parallel studies will investigate conditions that optimize robust adult learning and generalization to new talkers and words. Results from this research will contribute to the understanding of the limits of adult perceptual plasticity and the nature of language learning. It also will inform the development of effective and engaging next-generation educational tools.