Abstract

A fundamental mystery in music cognition concerns whether and how the human brain can develop expectations and preferences for events in the auditory environment. My thesis uses behavioral and electrophysiological methods to investigate the learning of a novel system of musical sounds. We design a new musical system based on the Bohlen-Pierce scale, a microtonal scale tuned differently from the traditional Western musical scale. Chord progressions and melodies were composed in this scale as legal exemplars of two finite-state grammars. In a series of behavioral studies, participants were presented with melodies in one of the two grammars, followed by several tests assessing grammar-learning, sensitivity to frequency of occurrence, and preference for melodies. Results demonstrate that given exposure to a small number of melodies, listeners recognized and preferred melodies they had heard, but when exposed to a sufficiently large set of melodies, listeners were able to learn the underlying statistical regularities of their given grammar. These effects were influenced by pshychoacoustic and statistical properties of the exposure, and were replicable with transposed melodies and for scales with different harmonies. Electrophysiological recordings (Event-Related Potentials) in response to chords in the new musical system revealed two components of cortical activity which are sensitive to the probability of occurrence and the amount of exposure of sounds in the musical context. We conclude that the human brain can rapidly acquire various structural and statistical aspects of sounds, and that neural mechanisms subserving statistical learning may be vital to music as well as other cognitive and perceptual functions more generally.

Dedication

To Mom and Dad

Acknowledgements

This dissertation would not have been possible without the guidance and support of various individuals from within and beyond UC Berkeley. I would like to thank David Wessel, my advisor and dissertation chair, for his unfailing support, expertise, and advice along the way. I would like to thank my advisor Erv Hafter for his kindness and support, and for his lessons in science and life. I express sincere gratitude to Bob Knight for his expertise in neuroscience as well as his kind advice and trust. To Carla Hudson Kam and the Language and Learning Lab (especially Amy Finn, Whitney Goodrich, and Tim Beyer) I am most indebted for advice, stimulating conversations, and moral support. For helpful discussions and technical support I thank Edmund Campion and the Center for New Music and Audio Technologies (especially Michael Zbyzynski, John MacCallum, Aaron Einbond, Brian Vogel, and Peter Kassakian), the Auditory Perception lab (Anne-Marie “Nannick” Bonnel, Tassos Sarampalis, Bernhard Seeber, and Andy Schmeder) and the Knight lab (Mark Kishiyama, Christina Karns, Cathrine Dam). I am extremely grateful to Tom Wickens for his generosity with help on statistical methods and especially with my use of his lab space. I owe a big thank you to all my research assistants over the past few years for their great productivity, intelligence, and good cheer: Elaine Wu, Pearl Chen, Tiffany Day, Judy Wang, Joann Chang, Young Lee, Jorge Duque, Johannes Sommer, Shaochen Wu, and Charles Li. I also thank Chris Lucas for help on implementing finite-state grammars, the Robertson lab (Ani Flevaris, Ayelet Landau, Joe Brooks) for helpful discussions on ERP methods, and Bill Prinzmetal for his kindness and support. I thank Carol Krumhansl at Cornell University for helpful advice on the design of the artificial musical system, and Marty Woldorff at Duke University for advising my undergraduate thesis, which was a vital precursor to the present dissertation. I thank the UC Berkeley Psychology Department, the Academic Senate, and NINDS for financial support. Finally, I would like to thank Wai-Po and Rebecca Loui.