Computational Modeling of Rhythm Perception and the Role of Enculturation

Abstract

Abstract This chapter compares a variety of computational models of rhythm perception and discusses them in three sections, each focusing on one of various different theoretical perspectives that exist in cognitive modeling, namely cognitivism, embodied cognition, and predictive processing. The different perspectives suggest different computational modeling techniques, which this chapter uses to differentiate models of rhythm perception. Cognitivism most naturally accommodates rule-based models, coupled oscillation models use mathematical tools associated with embodied cognition, and probabilistic generative models are consistent with predictive processing theories of cognition. Each section provides a short description of a theoretical perspective, followed by a discussion of rhythm perception models consistent with that perspective. Furthermore, the chapter draws attention to the influence that Western music theory may have had on models and theories of rhythm perception. This potential influence is of interest because rhythm perception is thought to be shaped by the history of experiences and activities of listeners, enabling the culture in which a listener is embedded to influence their perception. The chapter briefly reviews what effects this influence may have on rhythm perception, suggesting the need for modeling enculturated rather than “universal” listeners. Throughout, the chapter notes that rule-based models do not take previous experiences and activities of listeners into account, while some coupled oscillation models and probabilistic generative models, computational paradigms that gained popularity more recently, do, albeit to varying degrees. Additionally, probabilistic generative models, consistent with predictive processing, suggest a normative explanation of how previous experiences and activities shape perception.