Abstract Humans can rapidly regulate actions according to updated demands of the environment. A key component of action regulation is action inhibition, the failure of which contributes to various neuropsychiatric diseases, such as Parkinson’s disease (PD), obsessive compulsive disorder and Tourette syndrome. Action inhibition occurs in at least 3 ways: (i) action selection – selecting one action requires suppressing alternative motor plans, (ii) outright stopping – inhibiting a response when it is rendered inappropriate and (iii) action switching – change action plans in response to environmental changes. Despite the extensive effort to understand how the brain selects, stops and switches actions, the mechanism underlying these action regulation functions, and more importantly, how they inter-relate remain elusive. Part of this challenge lies in the fact that studies rarely explore, characterize, and investigate these functions together, making it difficult to develop a unified theory that explains the computational aspects of action regulation. The current proposal aims to advance our understanding by developing a neurocomputational model that, unlike prior models, integrates information from multiple sources (e.g., value of targets, cost for changing an action, contextual information) and predicts both kinematics of motor behavior and the underpinning neural mechanisms across 3 distinct types of action regulation. We will directly evaluate model predictions with intracranial recordings in patient volunteers undergoing deep brain stimulation implantation surgeries. These surgeries provide a unique opportunity to obtain multi-focal cortical and basal ganglia (BG) recordings with high temporal and spectral resolution and spatial specificity across the three action regulation tasks. The overarching goal will be achieved through three aims. In Aim 1, we will collect behavioral data from PD patients and aged-match neurotypical participants performing tasks that involves selecting, stopping and switching reaching actions. The motor behavior of the neurotypical group will be used to develop a neurocomputational model that simulate the fronto-BG circuits in action regulation. Then, we will assess how specific changes on the neural mechanisms of the model architecture predict the motor behavior of the PD patients. In Aim 2, we will evaluate the model predictions about the mechanisms of action selection relative to stopping by recording neural activity from PD patients who undergo surgery for DBS implantation of the subthalamic nucleus (STN). Neural recordings will be collected without and with temporally and spatially precise subthalamic nucleus (STN) stimulation to investigate the causal role of STN in action selection. In Aim 3, we will evaluate the model predictions about the mechanisms for switching actions by recording neural activity from PD patients with the STN stimulation off and on. Overall, successful completion will provide a unified theory of action regulation in the human brain, with both behavioral and physiological validation, opening new avenues on improving the effectiveness of neuromodulation with DBS and other neurorestorative therapies.

Project narrative How people select actions, cancel unwanted or inappropriate actions and switch actions in response to new incoming information is a fundamental neurobiological question that is of high impact for understanding how the human brain functions. Many neurological and psychiatric disorders, such as Parkinson’s disease (PD) and obsessive compulsive disorder (OCD), point to impaired brain circuitry responsible for these three suppression functions. The proposed study will develop a neurocomputational model to better understand these mechanisms and evaluate these models with direct recordings from patient volunteers during neurosurgical procedures.