Postdoctoral Associate Laboratory of Dr. Joshua T. Dudman HHMI Janelia Research Campus
Abstract: Deciding what actions to perform is central to how our inner selves interact with the outer world, a reality that is often not fully realized until these abilities are lost. The basal ganglia, a subcortical structure that consists of two opposing pathways, has been proposed to be responsible for encoding the reinforcement, valuation and selection of actions. To better understand this interface of cognition and movement in this region, we trained mice to make goal-oriented reaches while delivering cell-type-specific, optogenetic stimulation during only the fastest or slowest reaches. Contrary to prevailing notions of basal ganglia function, we found that biasing the activity of either pathway positively or negatively reinforced whatever action was stimulated. The direction of the reinforcement was determined by which pathway was stimulated and was independent of the valence of the action itself, implying a role in assessing value but not action selection.
Additionally we addressed the problem that despite most decisions existing on a spectrum of options (including the velocity of a movement), most neuronal decision models only take into account binary choices (left or right, stop or start). To address this issue, we combined computational modeling with simultaneous activity recordings dozens of neurons to create a mechanism that explain how the neural circuits make decisions along a continuous spectrum.
Finally, we applied dynamical systems approaches to analyze the information content of the basal ganglia. We found a distributed representation of movement that could predict actoin initiation, amplitude and duration over 100ms befor the behavior was performed. These experiments shed new light on how decisions are made, indicate a greater role for the basal ganglia in motor control, and suggest that the learning and selection of how to behave occurs in a hierarchical system throughout the brain. Future work, including the use of 1000 channel electrodes to thoroughly characterize the activity of multiple brain areas simultaenously, will test and build upon these models of neural control of behavior.