Neural and computational mechanisms of multi-option choice in prefrontal cortex - The goal of this proposal is to identify the computations and neural mechanisms for economic decisions between many options, using a nonhuman primate model. Multi-option choices are ubiquitous in natural settings but poorly understood. As a result, we lack basic knowledge of how the brain organizes real-world appetitive behaviors or why these behaviors change in neuropsychiatric disease. The focus of the proposal is how multiple options are compared, an essential computation that cannot be inferred from knowledge of simpler two-option choices. We frame the studies in terms of a two-stage, integrate-to-threshold decision mechanism: an input stage assigns values to each option, and an integration stage accumulates value signals for each option up to a bound. Our overarching question is where and how the options are compared to each other: at the input stage, at the integration stage, or both. Additionally, we seek to identify the functional form of these comparisons and their capacity limits when many options are available. Because existing data and models alone cannot answer these questions, we will use a tandem approach, first making cell-level observations of neural activity in a nonhuman primate model of multi-option choice, and then using these observations to constrain the specification of integrate-to-threshold models. In Aim 1, we record neural populations in the primate ventromedial and orbital prefrontal cortex (vmPFC and OFC), two regions consistent with the input stage in integrate-to-threshold models. The objective is to identify the broad class of functions that neurons use to compare decision options, i.e., to assign a subjective value to a given option in relation to the others available. In Aim 2, we identify the causal contribution of vmPFC and OFC to decision behavior, using intracranial microstimulation to disrupt value-coding activity in each region. Aim 3 focuses on comparisons at the level of integrated evidence using integrate-to-threshold models. Here we will fit a series of models in which we constrain the value comparison functions used at the input stage and ask how these constraints shape the strength of comparisons computed at the integration stage. Critically, a subset of these models will assume input functions derived from the vmPFC and OFC recordings in Aim 1. By using brain- derived constraints, these models are expected to give insights into integration mechanisms that could plausibly be realized in neural circuits. At their conclusion, these studies will provide novel empirical evidence showing how options are compared at early decision stages and new computational insights into how options are compared at later stages. By using a nonhuman primate model and a naturalistic behavioral task, the results are expected to have high translational potential. The proposed studies are innovative because they go beyond the binary choice frameworks that dominate the field and will result in new models that can be used in both monkeys and humans. And the findings are expected to be significant because they will provide new knowledge of how the primate prefrontal cortex organizes a complex form of everyday decision-making.
