Which Timbral Features Granger-Cause Colour Associations to Music?

Sensory information processing is inherently multimodal. An organism normally perceives the environment using all its senses simultaneously. Crossmodal correspondence might take place at any stage of neural processing (Spence 2011; Deroy & Spence 2016). A wealth of studies have has provided evidence that many non-arbitrary correspondences exist between auditory and visual stimulus features (Bresin 2005; Palmer et al. 2013; Whiteford et al. 2018). However, few studies take electroacoustic music compositions as substrate for stimuli, despite the great variations of timbre within and across such works. Time series analysis in music perception and cognition (Schubert 2001; Pearce 2011) has gained traction. This paper describes a method to evaluate audio-to-visual correspondences by analysing large multivariate time series with sonic features, extracted computationally, and perceptual visual colour associations, gathered empirically.