Decision making is a fundamental cognitive process, and many decisions are based on gradually accumulated evidence. Thus, it is critical to understand the mechanistic basis underlying this accumulation process. Traditional models of evidence accumulation are based on low-dimensional attractors where individual neurons show ramping activity throughout a trial. However, an increasing number of studies have observed choice-selective sequences in their neural recordings, in which neurons fire transiently and sequentially with the subset of neurons that fires indicative of the animal’s choice. Similar sequences have been observed in other memory and decision-making tasks, suggesting sequences are a fundamental form of neural dynamics that are inherently different from the persistent dynamics predicted by canonical models. To address the gap between classic models and emerging data, I will first develop two novel neural circuit models that accumulate evidence through sequences. The first will be a position-gated, bump attractor, where the set of active neurons (“location” of activity in the population) encodes position along one axis and accumulated evidence along the other so that evidence is encoded non-monotonically. In contrast to this location-based model, the second will consist of two mutually inhibitory chains, where the firing rate (“amplitude”) of the active neurons encodes evidence, so that evidence is encoded monotonically. Thus, these two novel models propose two alternative mechanisms for the accumulation of evidence through sequences, which are distinguished by their predictions about how evidence is encoded. Model predictions for the encoding of single neurons and the geometric structure of the whole-population code will be tested by comparison to an extensive set of previously collected neural activity during a navigation-based, accumulation of evidence task, which demonstrate choice-selective sequences across the brain, including in the hippocampus, visual cortex, anterior cingulate cortex, and striatum. Thus, this project will address two key questions in our understanding of decision-making, how evidence is accumulated through sequences and where these mechanisms are present in the brain, by proposing novel circuit mechanisms and analyzing neural data throughout the brain. The fellowship training plan equips me to explore these questions because I have chosen one experimental sponsor whose expertise lies in the systems neuroscience of reward learning and decision making and one computational sponsor with expertise in neural integrator models. Based at an institution at the leading edge of both experimental and computational neuroscience, I will have access to the resources to expand my neuroscience domain knowledge while gaining skills in computational neuroscience specific to modeling neural integration and neural encoding. Overall, this project will develop two novel paradigms for the accumulation of evidence, based on sequential neural activity, that elucidate differences in the underlying mechanisms across brain regions, suggesting how decision making may be coordinated across the brain.

PROJECT NARRATIVE Accumulating evidence over time to select between alternative options is a fundamental process in decision making, but canonical models suggest that neural signatures of this process should show persistent, ramping activity, unlike the choice-selective, sequential neural firing commonly observed throughout cortical and subcortical areas. This project will develop new computational circuit models that suggest different underlying mechanisms for accumulating evidence through sequences and analyze the correspondence between these novel models and neural recordings in different brain regions at the individual neuron and population levels during the same navigation-based, accumulation of evidence task. Successful completion of this work will lead to the development of novel models which propose yet unexplored mechanisms that may underlie decision making and advance our understanding of how decision making processes are distributed and coordinated across the brain.