Brain Plasticity in Speech Training in Native English Speakers Learning Mandarin Tones A Thesis SUBMITTED TO THE FACULTY OF UNIVERSITY OF MINNESOTA BY Christina Carolyn Heinzen IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS

consulting service about training methods. Many thanks to my advisor Dr. Zhang for his hard work and mentorship, and to my undergraduate assistants, Jiawei Li and Chloe Laabs, for their help with data collection. Additional thanks to my long-suffering and ever-supportive friends and family, who provided much-needed encouragement. ii Dedication This thesis is dedicated to all those hard-working individuals who, like me, struggled to learn a second language in adulthood. iii Abstract The current study employed behavioral and event-related potential (ERP) measures to investigate brain plasticity associated with second-language (L2) phonetic learning based on an adaptive computer training program. The program utilized the acoustic characteristics of Infant-Directed Speech (IDS) to train monolingual American English-speaking listeners to perceive Mandarin lexical tones. Behavioral identification and discrimination tasks were conducted using naturally recorded speech, carefully controlled synthetic speech, and non-speech control stimuli. The ERP experiments were conducted with selected synthetic speech stimuli in a passive listening oddball paradigm. Identical pre-and post-tests were administered on nine adult listeners, who completed two-to-three hours of perceptual training. The perceptual training sessions used pair-wise lexical tone identification, and progressed through seven levels of difficulty for each tone pair. The levels of difficulty included progression in speaker variability from one to four speakers and progression through four levels of acoustic exaggeration of duration, pitch range, and pitch contour. Behavioral results for the natural speech stimuli revealed significant training-induced improvement in identification of Tones 1, 3, and 4. Improvements in identification of Tone 4 generalized to novel stimuli as well. Additionally, comparison between discrimination of across-category and within-category stimulus pairs taken from a synthetic continuum revealed a training-induced shift toward more native-like categorical perception of the Mandarin lexical tones. Analysis of the Mismatch Negativity (MMN) responses in the ERP data revealed increased amplitude and iv decreased latency for pre-attentive processing of across-category discrimination as a result of training. There were also laterality changes in the MMN responses to the non-speech control stimuli, which could reflect reallocation of brain resources in processing pitch patterns for the across-category lexical tone contrast. Overall, the results support the use of IDS characteristics in training non-native speech contrasts and provide impetus for further research. As technology continues to advance, we find ourselves in a world that is continually shrinking. In this increasingly globalized world, where it is possible to communicate with people on the other side of the globe in a mere fraction of …