Robust Encoding of Abstract Rules by Distinct Neuronal Populations in Primate Visual Cortex

It is widely known that neural activity in sensory representations is modulated by cognitive factors such as attention, reward value and working memory. In such cases, sensory responses are found to reflect a selection of the specific sensory information needed to achieve behavioral goals. In contrast, more abstract behavioral constraints that do not involve stimulus selection, such as task rules, are thought to be encoded by neurons at later stages. We show that information about abstract rules is encoded by neurons in primate visual cortex in the absence of sensory stimulation. Furthermore, we show that rule information is greatest among neurons with the least visual activity and the weakest coupling to local neuronal networks. Our results identify rule-specific signals within a sensory representation and suggest that distinct mechanisms exist there for mapping rule information onto sensory guided decisions.