Adaptive oscillators provide a hard-coded Bayesian mechanism for rhythmic inference

Abstract

Bayesian theories of perception suggest that the human brain internalizes a model of environmental patterns to reduce sensory noise and improve stimulus processing. The internalization of external regularities is particularly manifest in the time domain: humans excel at predictively synchronizing their behavior with external rhythms, as in dance or music performance. The neural processes underlying rhythmic inferences are debated: whether predictive perception relies on high-level generative models or whether it can readily be implemented locally by hard-coded intrinsic oscillators synchronizing to rhythmic input remains unclear. Here, we propose that these seemingly antagonistic accounts can be conceptually reconciled. In this view, neural oscillators may constitute hard-coded physiological priors - in a Bayesian sense - that reduce temporal uncertainty and facilitate the predictive processing of noisy rhythms. To test this, we asked human participants to track pseudo-rhythmic tone sequences and assess whether the final tone was early or late. Using a Bayesian model, we account for various aspects of participants performance and demonstrate that the classical distinction between absolute and relative mechanisms can be unified under this framework. Next, using a dynamical systems perspective, we successfully model this behavior using an adaptive frequency oscillator which adjusts its spontaneous frequency based on the rate of stimuli. This model better reflects human behavior than a canonical nonlinear oscillator and a predictive ramping model, both widely used for temporal estimation and prediction. Our findings suggest that an oscillator may be considered useful as a potential heuristic for a rhythmic prior in the Bayesian sense. Together, the results show that adaptive oscillators provide an elegant and biologically plausible means to subserve (bayesian) rhythmic inference, thereby reconciling numerous empirical observations and a priori incompatible frameworks for temporal inferential processes.