Exploring the perception of expressivity and interaction within musical dyads

Abstract

Bodily gestures in music performance play an important role in the perception and appreciation of music by the audience. Performance variations can be identified by observers even when no auditory cues are available; visual kinematic information has been demonstrated to be crucial in identifying expressive intentions. The current thesis explores violin dyads performing an unfamiliar song. High quality optical motion capture was employed to record full body movement. We applied the standard paradigm to study the perception of expressive performance. Our aim was to predict perceptual ratings of expressivity from movement computations. That is, to deduce secondary aspects of musical gestures, namely intentions, from primary aspects, namely physical movement. Our hypotheses were that perceived expressivity is dominated by vision and that ancillary gestures carry significant cues for observers. For this purpose, a web-based perceptual experiment was conducted using short duration stimuli that were rendered as point light representations. Sound and vision, vision only, and sound only stimuli were employed to explore interactions across different modalities. Preliminary movement analysis showed that the musicians performed with greater amounts of kinetic energy in the more exaggerated expressive conditions. We extracted low level kinematic features based on instantaneous velocities of the markers of interest. We applied principal component analysis on the motion capture timeseries data, and we performed multiple linear regression and linear discriminant analysis to assess our hypotheses. We extend previous findings about visual perception in dyadic context, and we provide an account of idiosyncratic movements in violin performance.