Descripción del proyecto
Decision-making behaviors often occur in the absence of clear instruction to guide action. Instead, theories and experiments have predicted that the brain must compute a decision-value based on past experience to select the best action. This implies that the action with the highest subjective value should always be chosen. However, behavior is often stochastic with variability from trial-to-trial. To resolve this long-standing paradox, MOTORHEAD will take full advantage of state-of-the-art in vivo neuronal recordings and computational methods in behaving rodent to bridge for the first time the gap between deterministic decision-signal and stochastic motor commands, achieving thus an unprecedented level of understanding of these unpredictable behaviors. Indeed, despite decade of intensive work, key questions remain unexplored: i) How such a deterministic decision signal is maintained without necessarily causing movement? ii) And how it is then converted to a final motor command with trial-by-trial variability? Here, we hypothesize that these two operations occur in higher-order motor areas, and more particularly across recurrent cortical layers of the secondary motor cortex of rodents. Specifically, we posit that: i) Distinct populations of layer (L) 5 pyramidal neurons (PNs) generate specific movement according to the decision statistics provided by L2/3 PNs. Specific attractor architectures, with different stability to noise perturbation, could cause the system to behave more or less randomly. ii) This top-down excitation could be dynamically gated by bottom-up plasticity forces from reward-related structures, which modulate decision-value to account for past choice outcome, notably when the action no longer generates the expected outcome. To achieve this breakthrough, we propose an ambitious system neuroscience approach, at high spatial and temporal resolution, to illuminate the cellular principles underlying the control and transformation of decision variable.