Abstraction of reward context facilitates relative reward coding in neural populations of the anterior cingulate cortex

The anterior cingulate cortex (ACC) is believed to be involved in many cognitive processes, including linking goals to actions and tracking decision-relevant contextual information. ACC neurons robustly encode expected outcomes, but how this relates to putative functions of ACC remains unknown. Here, we approach this question from the perspective of population codes by analyzing neural spiking data in the ventral and dorsal banks of the ACC in monkeys trained to perform a stimulus-motor mapping task. We found that neural populations favor a representational geometry that emphasizes contextual information, while facilitating the independent, abstract representation of multiple task-relevant variables. In addition, trial outcomes were primarily encoded relative to task context, suggesting that the population structures we observe could be a mechanism allowing feedback to be interpreted in a context-dependent manner. Together, our results point to a prominent role for ACC in context-setting and relative interpretation of outcomes, facilitated by abstract, or "untangled," representations of task variables. Author SummaryThe ability to interpret events in light of the current context is a critical facet of higher-order cognition. The anterior cingulate cortex is suggested to be important for tracking information about current contexts, while alternate views hold that its function is more related to the motor system and linking goals to appropriate motor responses. Here, we evaluated these two possibilities by recording anterior cingulate neurons from monkeys performing a stimulus-motor mapping task in which compound cues both defined the current reward context and instructed appropriate motor responses. By analyzing geometric properties of neural population activity, we found that the ACC prioritized context information, representing it as a dominant, abstract concept. Ensuing trial outcomes were then coded relative to these contexts, suggesting an important role for these representations in context-dependent evaluation. Such mechanisms may be critical for the abstract reasoning and generalization characteristic of biological intelligence.