Learning and decision-making are driven by expectations of future outcomes. Three key parameters determining the valuation of future outcomes are 1) ?how much? to expect, 2) ?when? to expect it, and 3) ?what? to expect (ie, Outcome Prediction). However, how outcome prediction is generated by the brain in response to predictive cues is poorly understood. Exemplifying the when of outcome-prediction is so-called ?reward timing? activity in the primary visual cortex (VC), which emerges in VC when visual stimuli are behaviorally conditioned with delayed water reward. Previously, we have demonstrated that this timing activity is generated within VC itself and requires basal forebrain cholinergic innervation to be formed. We have also demonstrated that this activity informs on the timing of visually-cued actions. Indeed, the dorsal striatum (DS) is VC's direct downstream motor-related target, and it is also observed in pilot data to expresses this activity. Together, these observations make the visual corticostriatal circuit (VCDS) a powerful system to address how outcome prediction can be learned and reported neurally. Combined with our computational model of how outcome prediction signaling could be learned by reinforcement signaling within VCDS, these observations well motivate our research into how VCDS circuitry produces outcome prediction signals, how cholinergic signaling teaches this circuit to learn outcome predictive signaling, and whether predictive signaling in VCDS informs decision-making behavior. Whether appetitive (Aim1a) and aversive (Aim1b) conditioning leads to the visual corticostriatal circuit learning to produce outcome prediction signals is unknown, though pilot data indicates it is. Testing predictions from our formal model, selectively perturbing inhibitory circuit elements will assess whether VC is a site sourcing predictive signaling to DS (Aim1c). Pilot Ca2+ imaging of cholinergic fibers within VC indicates that reward, as well as punishment is reported to VC (Aim2a) in keeping with its purported role as a teaching signal, but raising the possibility that outcome valence is learned downstream in DS (Aim2b). Therefore, the degree of cholinergic activation within VC may serve to teach VC to express and source to DS signals predicting the time and magnitude of expected outcomes (Aim2c), while DS may serve as a site associating those predictive signals with their appropriate reward-seeking/punishment avoiding behaviors. The ability to optogenetically mimic outcome signaling affords a means to test whether learned outcome prediction signaling in VCDS informs decision-making: By instilling fictive reward expectancies atop behaviorally conditioned reward expectancies of otherwise equal value, outcome prediction signaling in VCDS can be shown to impact future decision making (Aim3a&b). Observations made here will advance an understanding of the mechanisms?impaired in many cognitive diseases?of how the behavioral meaning of sensory information is learned in order to remember past experiences and inform decision making.
Learning and decision making are driven by expectations of future outcomes, yet how outcome prediction is generated by the brain to learn and inform behavior is poorly understood. We propose to investigate this issue by using a model system for the study of reinforcement-based learning?the visual corticostriatal circuit?which affords control over inputs predictive of future outcomes, as well as inputs conveying the occurrence, magnitude, and valence of outcomes themselves. The goals of the work are to: 1) characterize corticostriatal signaling of outcome prediction and the circuitry producing it, 2) measure cholinergic and dopaminergic signaling of rewards and punishments and establish how reinforcement signaling teaches corticostriatal circuitry to report outcome prediction, and 3) demonstrate that this outcome prediction signal causally informs decision making.
