Mechanisms of Valuation: Encoding of Outcome Variables in Orbitofrontal Cortex and Ventral Striatum

Abstract

Adaptive decision making requires that we consider not only the inherent value of our options but also more specific features of those options, which can be used to compute the current value of each choice in a dynamic environment. General value information is conveyed via `model-free' signaling, whereas details of outcomes that are independent of value are conveyed via `model-based' representations of outcomes. Orbitofrontal cortex (OFC) and ventral striatum (VS), which function as part of the reward system, have been implicated in value-guided behaviors, but their specific contributions to model-free and model-based signaling remains elusive. Several studies suggest that OFC is not critical for distinguishing differentially valued outcomes of a common currency. Instead, encoding of specific features of outcomes in OFC appears to provide the flexibility required for advantageous choice selection in several value-guided behaviors. However, VS has been shown to be essential when either value or specific feature information is necessary for adaptive behavior. In the following set of experiments, we tested the hypothesis that OFC signals model-based information, which is incorporated with model-free signals in VS. To isolate model-free and model-based signals, we independently manipulated size (value) and flavor (specific feature) of rewards while recording single units in the rat OFC (Experiment 1) or VS (Experiment 2). Our data provide evidence for model-based signals in OFC and evidence for a hybrid of model-based and model-free signals in VS. Whereas OFC lesions disrupted model-based representations of flavor in VS, they did not eliminate all model-based signaling. We therefore conclude that OFC provides some, but not all, of the model-based information in VS.