Reinforcement regulates timing variability in thalamus

Abstract

the project. J.W. and E.H. collected the main behavioral and electrophysiology data. J.W. analyzed the data and developed the model. E.H. played a major role in data analysis. J.W and M.J. interpreted the data with contribution from E.H. and N.M.. N.M. performed and analyzed the control experiment in a third monkey, and highlighted the need for additional validation analyses. A.A. conducted the human psychophysics experiments. M.J. supervised the project. All authors contributed to the writing of the manuscript. AbstractMotor learning reduces variability and motor variability facilitates exploratory learning. This paradoxical relationship has made it challenging to tease apart behavioral and neural signatures of variability that degrade performance from those that improve it. To tackle this question, we analyzed behavioral variability and its correlates across populations of neurons in the premotor cortex, thalamus and caudate within the cortico-basal ganglia circuits during a flexible motor timing task. Behavioral variability was comprised of two opposing processes: a slow drift in memory that caused effector and interval specific variability, and a fast reinforcement learning process that countered memory drifts in a context-specific manner. In all three brain regions, memory drifts were accompanied by context-specific fluctuations of neural activity along behaviorally-relevant dimensions. However, only in thalamus was drift-related neural variability regulated by reinforcement. Previously, we found that thalamus provides a speed command to control response dynamics in cortex and caudate responsible for flexible motor timing. Our current findings indicate that the nervous system uses reinforcement to regulate the variability of the speed command in thalamus in order to calibrate memory in the presence of inherent slow drifts. More generally, our work provides an example of reinforcement acting upon neural variability to improve behavioral performance.

The Central Thalamus: Gatekeeper or Processing Hub?

“…There are several reasons why this is a useful paradigm for investigating information integration in the central thalamus. First, multiple lines of evidence across species suggest a central role for the central thalamus in timing movements (Tanaka, 2006;Lusk et al, 2020;Wang et al, 2020). Second, the authors can leverage their own prior investigations of signals in cerebellar nuclei and striatum (Ohmae et al, 2013;Kameda et al, 2019) to determine which components of thalamic activity might rely on these dominant inputs to the central thalamus.…”

A proxy measure of striatal dopamine predicts individual differences in temporal precision

“…Importantly, Cheng et al (2007) highlighted that the transition to a dopamine-insensitive state was similar to general observations of striatal neurons becoming silent once a reward becomes predictable with training (Schultz, 1998). Similarly, Wang et al (2020) recently suggested that the noise in corticostriatal circuits that underlies timing variability is subject to adjustments through reinforcement learning. This overlap between learning and timing may help to reconcile our seemingly discrepant results.…”

Variance and Scale-Free Properties of Resting-State BOLD Signal Alter after Fear Memory Acquisition and Extinction

“…At the same time, all BN-ROIs overlapped with FEN and TRC masks showed a gradual increase only in the variance of the averaged BOLD signal. Such multidirectional changes in variance may indicate different information processing in the cortical and subcortical areas (Wang et al, 2019).…”