A listener’s propensity to perceive affect as expressed by music can arise from factors such as acoustic features and culturally learned expectations. Studies investigating the link between musical flow and perceived affective content by means of continuous response measures and a 2-dimensional circumplex framework of affect (i.e., arousal and valence) have given positive results. For example, time series models of perceived arousal in response to Western classical and electroacoustic music reveal a significant predictive influence of acoustic parameters such as intensity and spectral flatness. Acoustic parameters generally provide weaker models of perceived valence. Here we test the hypothesis that a continuous measure of musical engagement can be a significant predictor of perceived arousal and perceived valence, and will enhance time series models of affect based on acoustic parameters alone. Thirty-five nonmusicians continuously rated their level of engagement while listening to 5 Western classical and electroacoustic music excerpts. Grand unweighted mean engagement time series for each piece from all 35 participants were used to model continuous-response time series of perceived arousal and perceived valence. The hypothesis was partially supported: in univariate autoregressive analyses, 1 of the valence and 2 of the arousal models were strongly improved by adding engagement as a predictor; and in a further 2 of each, engagement made a minor contribution. In the remaining 2 models of valence and 1 of arousal, engagement was not pertinent. In multivariate (vector autoregressive) models, relating simultaneously both arousal and valence to acoustic parameters, engagement had a role in every case. It is concluded that listener engagement can play a mediating role in perceived affective response to music.
- time series analysis