Memory-driven coding of objects in monkey parieto-frontal grasping network

Every goal-directed action begins with a computational challenge: translating what is seen into how to move. Yet kinematically precise movements can be generated in complete darkness, as in grasping familiar or just-seen objects relying on memory-based mechanisms that bridge perception and action. How does the brain transform object information into appropriate motor plans in the absence of vision? To address this question, we recorded single-unit activity (n = 551) from key parietal and frontal nodes of the macaque extended grasping network (AIP, F2, F5, F6) during both visually- and memory-guided grasping.Area AIP stands out as a central hub for memory-driven actions, with neurons exhibiting temporally stable, object-specific tuning from object presentation through grasp execution in darkness, indicating a parietal substrate that maintains object representation during memory-driven grasping. In contrast, premotor regions showed transient, phase-locked, context-dependent activity, consistent with flexible transformation of parietal information into motor output rather than its maintenance. At the cellular level, object stability in AIP was strongest among broad-spiking (excitatory-like) units, consistent with local recurrent dynamics able to stabilize the code and support its redistribution across the parieto-frontal network. Together, these findings outline a distributed parieto-frontal organization in which parietal stability and frontal flexibility jointly keep perception and action aligned, even when sensory input is unavailable.