Decision-making mechanisms underlying context-dependent object categorization

Visual recognition of objects is critical for guiding appropriate actions in diverse behavioral contexts. While its sensory mechanisms have been extensively studied, how decision-making processes operate during object categorization is poorly explored. Here, I present two findings made through a decision-theoretic approach. First, we recorded from an oculomotor area (LIP) while monkeys reported face category through eye movements. We found that neural activity reflects decision formation, but the patterns differ drastically from the evidence-accumulation signals reported previously in simpler sensory discrimination. The activity forms a non-linear population manifold that rotates across tasks, suggesting context-dependent mechanisms. Second, we found that changing task context behaviorally affects the accumulation of evidence during object categorization. When human observers switch between two categorization rules for the same face images, there is a brief reduction in evidence accumulation that recovers only after a few hundred milliseconds from stimulus presentation. Together, we argue that object recognition is mediated by intricate decision-making mechanisms that cannot be readily extrapolated from simpler perceptual tasks.