Introducing and testing a Bayesian emotions-in-music model (BEiMM): a cross-cultural investigation of Japanese and UK younger and older listeners

What is the basis for listeners to perceive music as expressive of emotions? For almost three decades, the cues-based lens model has been the primary method to predict perceived emotions from musical properties. This paper introduces a much-needed updated model that integrates listeners’ subjective conceptualisations with cues-based signification. Our Bayesian Emotions in Music Model (BEiMM) uses Bayes’ rule to predict emotion in music from the product of emotion concepts and expected relations between musical cues and emotion dimensions, moderated by characteristics of the musical genre. It acknowledges the role of active sense-making in affective responses and variations in emotion concepts participants may hold. A proof of concept is offered using a cross-cultural comparative study of listeners who have similar levels of familiarity with the presented stimuli. The results showed overlap and significant variations in emotion concepts and perceived emotions in music between Japanese and UK listeners, younger and older adults. The findings demonstrate the ability of BEiMM to accurately predict perceived emotion in music, and highlight the important role of listeners’ emotion concepts and their cross-cultural variation. Implications relate to the need to update how we understand emotions in music, with relevance for the modelling of emotions in other art and everyday domains.