Cingulate-centered flexible control: physiologic correlates and enhancement by internal capsule stimulation

The flexible deployment of cognitive control is essential for adaptive functioning in dynamic environments given limited cognitive resources. That flexibility depends on rapid detection and resolution of control-prediction errors (CPEs) when current demands diverge from the control plan. Deficits in control and control flexibility are common in psychiatric disorders, yet targeted interventions are limited by incomplete circuit-level understanding and limited means for modulating control circuits . We analyzed two intracranial electroencephalography datasets (one with brief internal capsule stimulation, ICS) to identify a human neurocomputational mechanism for CPE resolution and to test its modifiability. A third dataset of patients receiving internal capsule deep brain stimulation (IC DBS) assessed clinical relevance of modifying CPE-related processes. Phase-amplitude coupling (PAC) anchored to the θ phase of right rostral anterior cingulate cortex (rACC-R), especially θ-γ coupling between rACC-R and nodes of the cognitive control network (dorsolateral prefrontal cortex, dlPFC; dorsal ACC, dACC), was associated with faster CPE resolution. An adaptive drift-diffusion model indicated that ICS improves control flexibility specifically under high CPE, and mediation analyses showed that this behavioral improvement is mediated by CPE-dependent increases in rACC-R θ-centered PAC. In a psychiatric cohort (N=14; primarily treatment-resistant depression, TRD) with IC DBS, enhanced control flexibility, rather than CPE-independent general cognitive control, was strongly associated with clinical response (AUC = 0.90), suggesting both a behavioral flexibility index and rACC-R PAC as candidate biomarkers for DBS optimization. These findings identify a rACC-centered, θ phase-based coordination of the cognitive control network as a neurocomputational substrate of flexible control. They demonstrate that capsule stimulation selectively augments this substrate when flexibility is required, and establish flexibility, rather than general control, as the feature that tracks therapeutic benefit in TRD. Together, they suggest actionable biomarkers to guide, personalize, and potentially enable closed-loop neuromodulation for disorders marked by cognitive rigidity.