Marble melancholy: using crossmodal correspondences of shapes, materials, and music to predict music-induced emotions

Abstract

Introduction: Music is known to elicit strong emotions in listeners, and, if primed appropriately, can give rise to specific and observable crossmodal correspondences. This study aimed to assess two primary objectives: (1) identifying crossmodal correspondences emerging from music-induced emotions, and (2) examining the predictability of music-induced emotions based on the association of music with visual shapes and materials.

Methods: To achieve this, 176 participants were asked to associate visual shapes and materials with the emotion classes of the Geneva Music-Induced Affect Checklist scale (GEMIAC) elicited by a set of musical excerpts in an online experiment.

Results: Our findings reveal that music-induced emotions and their underlying core affect (i.e., valence and arousal) can be accurately predicted by the joint information of musical excerpt and features of visual shapes and materials associated with these music-induced emotions. Interestingly, valence and arousal induced by music have higher predictability than discrete GEMIAC emotions.

Discussion: These results demonstrate the relevance of crossmodal correspondences in studying music-induced emotions. The potential applications of these findings in the fields of sensory interactions design, multisensory experiences and art, as well as digital and sensory marketing are briefly discussed.

Keywords: crossmodal correspondences; machine learning; materials; music-induced emotions; random forests; sensory interactions; shapes.