PROJECT SUMMARY
The basal ganglia (BG), a set of highly interconnected nuclei deep in the brain, are critical to normal
motor and cognitive functioning in humans; diseases of the BG impair everyday life for people with
Parkinson’s disease (PD) and dystonia. Studies from non-human primates and rodents have suggested
that during well-learned motor sequences the BG send transition signals to cortex via thalamus to stop
the execution of one motor chunk and start the next. In this study, we will the hypothesis that the BG
send transition signals during speech production, one of the most complex discrete motor behaviors.
We will leverage the rare opportunity to record from human BG during deep brain stimulation (DBS)
surgeries. DBS brain region targets include the subthalamic nucleus (STN) and the globus pallidus
internus (GPi), both important nodes in the BG. Most previous studies analyzing these intraoperative
intracranial recordings have focused on the STN. This project will extend empirical evidence available
for the role of the GPi (the primary output nucleus of the BG) in human behavior. Analyzing high
spatiotemporal local field potentials from the GPi during a speech production task, will test the
hypothesis that the BG send transition signals at the end of each motor chunk in a sequence of well-
learned chunks (Aim 1). Preliminary data are concordant with prior hypotheses about the level at which
speech motor chunks are coded in the BG—at the syllable level, as opposed to at the phone or phrase
level. In Aim 2, we will advance the DIVA/GODIVA computational model of speech production with
three main areas of development: designing distinct STN and GPi modules, constraining the speech
motor chunk size of the BG according to Aim 1 findings, and modeling the BG learning the appropriate
sensorimotor context for each speech motor chunk.
In sum, this project proposes to bring together a) human speech production intracranial recordings of
the BG during deep brain stimulation implantation surgery from the Brain Modulation Lab (Mass
General Hospital) and b) speech computational modeling from the Speech Neuroscience Lab (Boston
University). These two areas will work synergistically in that human intracranial data is in need of
constrained hypotheses and speech computational modeling is in need of time-resolved empirical data
from the BG. The proposed project will contribute to the NIDCD’s mission to advance our understanding
of the neural basis of speech production by leveraging unique data and principled modeling. The
training strategy in this project is also uniquely promising to advance NIDCD’s mission to support cross-
disciplinary communication science research careers that bridge laboratory research and patient care;
my goal is to become an independent researcher in speech neuroscience.