DEEP BRAIN STIMULATION RESHAPES THALAMIC DYNAMICAL TIMESCALES DURING GOAL-DIRECTED MOTOR CONTROL

Deep brain stimulation (DBS) alleviates pathological motor symptoms in neurologically impaired subjects, yet its circuit-level mechanisms of action remain unclear. We hypothesized that cerebellum- and basal ganglia–receiving thalamic motor nuclei implement the functional factorization required by active inference: cerebellar-recipient motor thalamus (VLp-like domains) conveys predictive content, whereas basal ganglia–recipient motor thalamus (VLa/VA/VM-like domains) regulates precision via thalamocortical gain control. We tested this framework using dynamical and spectral analyses of human thalamic local field potentials (LFPs) recorded from contralateral ventral intermediate (VIM) and ventral oralis (VO) leads during three force-control paradigms. Trials were segmented into Plan, Delay, and Active epochs based on instruction and go cue onsets followed by force control. Intrinsic dynamical timescales were estimated using state-space models, yielding epoch- and stimulation-specific time constants (τ) related to spectral knee frequencies. Across all paradigms, VIM stimulation significantly reduced τ in both VO and VIM recordings relative to DBS-Off, indicating increased damping, whereas VO stimulation produced intermediate effects. Oscillatory amplitudes quantified via band-limited Hilbert envelopes showed selective modulation of frequency bands in VIM and VO, driven primarily by differences between rest and task engagement rather than between task epochs. Together, these findings indicate that DBS primarily reshapes intrinsic thalamic dynamics, with more limited, band-specific effects on oscillatory amplitude. The consistent reduction of thalamic time constants under VIM stimulation supports DBS as a control input that increases effective damping of belief dynamics, consistent with a precision-weighting mechanism that scales prediction’s influence on action selection and execution.