Name:Chihiro Honda

School/Affiliation:University at Buffalo, SUNY

Co-Authors:Anthony Nguyen, Peter Q. Pfordresher, and Chris McNorgan

Virtual or In-person:In-person

Abstract:

Previous studies reported different patterns of neural activations during the perception of familiar vs unfamiliar languages (Schlosser et al., 1998) as well as for different musical styles (Koelsch et al., 2000). However, Morrison and colleagues (2003) investigated the effects of music and language familiarity and expertise on brain activity using fMRI and concluded that listening to culturally different musical styles recruits similar neural resources. The current study utilizes machine learning techniques to investigate the organizational roles that musical experience, stimulus domain, and familiarity play on the neural processing dynamics of the auditory system. The dataset was from the study by Morrison and colleagues (2003), which consisted of anatomical and functional MRI images of 5 Western-trained violinists and 4 musically untrained individuals who listened to familiar and unfamiliar music (Western and Chinese) and to familiar and unfamiliar languages (English and Cantonese). We computed functional connectivity between each pair of the target regions and ran machine learning classifiers to identify the patterns of functional connections that predict the domain (music vs language) and familiarity (familiar vs unfamiliar) of stimuli and participants’ musical experience (trained vs untrained). Our models classified musical experience and the domain with high accuracy. However, our models failed to predict the familiarity factor. Thus, certain top-down factors (i.e., musical training) and bottom-up factors (i.e., acoustic structures in language vs music) seem to influence neural connectivity for auditory processing, but there may not be specific patterns of neural connectivity that distinguish familiar stimuli from unfamiliar ones in the auditory system.