Neural dynamics of decision-making and cognition

Our lab studies how the brain turns sensory input into the right action at the right time. We are guided by the belief that understanding how the brain works is the surest path to better interventions for people with disorders of the nervous system. We expect our work to inform translational efforts, from circuit-level therapies to brain-machine interfaces for patients with stroke, brain injury, and other neurological and psychiatric conditions. To get there, we combine large-scale electrophysiology, cell-type-specific tools, careful behavior, and computational modeling in macaque monkeys. Our work falls into three connected areas.

Decision-making

We ask how multiple brain areas cooperate to make flexible decisions. We train monkeys on rich tasks that pull perception apart from action. In one task, monkeys judge the dominant color of a red-green checkerboard and report it by reaching to a target whose location is randomized, which turns the decision into an exclusive-or (XOR) problem. Recording with Neuropixels probes in dorsolateral prefrontal cortex and dorsal premotor cortex, we find a division of labor. Prefrontal cortex carries the perceptual judgment and solves the XOR problem, and premotor cortex carries the final action. Adding delays lets us separate deliberation, working memory, and motor planning within the same trial. We interpret these data through a dynamical systems lens and build multi-area recurrent neural network models to ask what each area computes and how signals flow between them. We are also studying how neuromodulators such as norepinephrine and dopamine shape decisions through arousal and motivation.

Grasping

Picking up a cup looks effortless, but it requires the brain to coordinate where to reach, how to shape the hand, and how much force to use. We study how this skill emerges from activity in premotor and motor cortex. Monkeys reach for and lift objects that vary in location, size, shape, texture, and weight while we record from many neurons and track their hand and arm with markerless video tracking. We are mapping how dorsal premotor, ventral premotor, and primary motor cortex share information as a grasp unfolds, and when each object property becomes available to the brain. We are now beginning to test these interactions causally with optogenetics. This work speaks directly to restoring hand function after stroke and to building better brain-machine interfaces.

Bridging circuits and dynamics

Population dynamics tell us what a brain area computes, but not which neurons do the computing. We want to close that gap. We developed WaveMAP and PhysMAP, machine learning methods that classify cell types from their electrophysiological signatures. With colleagues at the Allen Institute, we are validating viral tools that use enhancers to target inhibitory neurons in the primate brain, which lets us opto-tag and manipulate specific cell classes during behavior. We are extending this with PV- and SST-specific constructs and an atlas of interneuron enhancers in macaque prefrontal cortex. Together, these let us ask how particular cell types shape the dynamics that drive decisions and movements, in both healthy and injured primate brains.