Distributed burst activity in the thalamocortical system encodes reward contingencies during learning

Neuronal bursts are distinct high-frequency firing patterns that are present ubiquitously throughout mammalian brain circuits. Although bursts are considered part of a universal neural code, the information they convey has long been a subject of debate. In this study, we investigated neuronal activity in simultaneously recorded regions of the thalamocortical system in freely moving mice as they learned stimulus-outcome associations in a go/no-go task. We discovered that, in parallel with learning, populations of neurons emerge in cortical, thalamic, and extrathalamic regions of the somatosensory system that encode task-relevant stimulus features via the presence or absence of bursts. These burst-coder neurons (BCNs) increase in number with task proficiency and exhibit burstiness that scales with stimulus valence rather than physical stimulus identity. Notably, BCNs consistently track stimulus-outcome associations—even after multiple rule switches—by inverting their burst encoding of the physical stimuli, indicating that burst coding is driven by outcome associations rather than by inherent stimulus properties. Although burst coding emerges throughout the thalamocortical system, only cortical units retain significant burst coding after devaluation, while other regions lose their discriminative burst patterns. Furthermore, decoding of stimulus properties and behavior achieves maximal accuracy when bursts or BCNs are used as input. Overall, these results provide direct experimental evidence linking neuronal bursting to learning, supporting a novel perspective of bursts as context encoders and teaching signals.