Actions are not mediated solely by cortical processes but rely on communication within basal ganglia-thalamocortical loops. Speech is one example, although how the basal ganglia participate in this uniquely human behavior is not clear. Recognizing that deep brain stimulation (DBS) surgery offers the only opportunity to directly measure neural activity in the human basal ganglia, we initiated studies to understand how speech information is encoded within the subthalamic nucleus (STN)-sensorimotor cortical network. Our team works with patients undergoing deep brain stimulation for movement disorders like Parkinson's disease and essential tremor. We established a novel experimental paradigm, where electrocorticography (ECoG) is recorded simultaneously with STN single unit activity and local field potentials (LFP), during DBS surgery in which patients are awake and speaking. We discovered that STN neuron activity is dynamic during speech production, exhibiting behaviorally-selective inhibition and excitation of separate populations of neurons. At the population level, we found that STN activity tracks with specific articulatory motor features and with gain adjustment in articulatory movements. In addition, our data suggest a role for the STN in speech planning, in that STN activity appears to be modulated prior to speech onset. The goal of this work is to build a neurobiological framework that allows the development of interventions to treat speech disorders involving the basal ganglia.