Independent Basal Ganglia Neural Populations Encode Speech Production and Ambient-Noise Levels

The function of the basal ganglia (BG), a set of nuclei deep in the brain, is still under debate despite their clinical relevance. Lines of evidence from non-human primate and human experiments suggest that one key causal contribution of the adult primate BG is invigoration of movement according to environmental and energetic constraints. Previous studies have come to this conclusion using primarily limb motor control tasks. We hypothesized BG are also important for invigorating the respiratory, vocal, and articulatory motor systems during speech production. To test this, we used a rare clinical opportunity to record local field potentials-broadband gamma activity (LFP-BGA) and single unit (SU) activity from human basal ganglia during deep brain stimulation implant surgeries. Participants repeated sentences aloud in the operating room. We modulated environmental and energetic constraints of speech production by inducing the Lombard Effect, wherein people reliably increase the intensity of their voice in the presence of background noise. Participants increased voice intensity, pitch, vowel space size, and respiratory depth in the presence of ambient noise. ∼40% of LFP and SU sites were modulated by speech production. Comparing globus pallidus (GP) LFP-BGA, herein reported for the first time during speech, to subthalamic nucleus (STN) LFP-BGA revealed GP activated after STN. STN was more coincident with the speech preparatory period and GP more coincident with speech onset, suggesting differential roles in speech production. We found only weak evidence that STN or GP activity encodes the level of produced speech intensity. Instead, we found that LFP-BGA and SU-FR robustly track the ambient background noise levels. Linear decoding from SU-firing rate revealed we could infer the background noise levels with >75% accuracy in any given time point in the trial, including baseline. Importantly, the ambient-noise-modulated sites were an independent population from speech-modulated sites. We conclude that human BG encode the energetic constraints of communication. We theorize that the code is leveraged in downstream BG-cortical loop nodes to invigorate speech oro-motor gestures and co-speech manual limb gestures, both of which respond to ambient noise levels to maintain high-fidelity communication.