Motor and Predictive Processes in Auditory Beat and Rhythm Perception

Abstract

In this article, we review recent advances in research on rhythm and musical beat perception, focusing on the role of predictive processes in auditory motor interactions. We suggest that experimental evidence of the motor system's role in beat perception, including in passive listening, may be explained by the generation and maintenance of internal predictive models, concordant with the Active Inference framework of sensory processing. We highlight two complementary hypotheses for the neural underpinnings of rhythm perception: The Action Simulation for Auditory Prediction hypothesis (Patel and Iversen, 2014) and the Gradual Audiomotor Evolution hypothesis (Merchant and Honing, 2014) and review recent experimental progress supporting each of these hypotheses. While initial formulations of ASAP and GAE explain different aspects of beat-based timing-the involvement of motor structures in the absence of movement, and physical entrainment to an auditory beat respectively-we suggest that work under both hypotheses provide converging evidence toward understanding the predictive role of the motor system in the perception of rhythm, and the specific neural mechanisms involved. We discuss future experimental work necessary to further evaluate the causal neural mechanisms underlying beat and rhythm perception.

Expectancy-based rhythmic entrainment as continuous Bayesian inference

“…Predictive processing should be a natural modeling framework for understanding rhythmic expectation and entrainment [14][15][16]. However, existing predictive coding models that operate in continuous time are structured to perform inference based on continuous observation, characterizing prediction errors in terms of deviation between a true level of input and a mean expected level [17,18].…”

Mapping between sound, brain and behaviour: four-level framework for understanding rhythm processing in humans and non-human primates

“…Moreover, it is important to keep in mind that biomechanical constraints which limit movement at fast rates may not directly apply to the internal pulse representation. This is evidenced by the ability to internally represent faster pulses that can be directly executed through stable movement [49]. Yet, such considerations do not contradict the wide range of evidence suggesting that the internal representation of a metric pulse might be implemented within neural circuits involved in motor control [40,50].…”

“…Hence, real-time mechanisms must be used to maintain the precise one-to-one temporal coordination between the input and internal pulse representation (and consequently between the input and overt behaviours that use the internal pulse as a temporal reference). Studies of externally paced finger tapping responses (the sensorimotor synchronization paradigm) have revealed that these mechanisms include reactive error correction processes entailing phase and period correction [49]. Moreover, additional online mechanisms include anticipatory processes that allow the timing of upcoming sensory events to be predicted during ongoing tempo changes [1,130].…”

Vocal learning as a preadaptation for the evolution of human beat perception and synchronization

“…[31][32][33][34]). Overall, the networks involved in beat perception without movement and in BPS have a great deal of overlap [35,36], and there is growing interest in the idea that motor system activity plays a causal role in predicting the timing of beats even when humans do not move to the beat [37][38][39][40][41].…”