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

Parieto-frontal networks are thought to play a crucial role in coordinating the visuomotor transformations required to interact with objects and other subjects in the environment. In particular, several parietal and frontal areas have been shown to constitute an “extended grasping network” for transforming the physical features of graspable objects into the kinematic parameters of the hand and arm for reaching-grasping actions, which are continuously updated based on visual feedback. However, objects with known features can also be grasped in the dark, relying on memory- rather than sensory-motor transformation. Previous evidence suggests a role for posterior parietal cortex in memory-guided hand manipulation, but the relative contribution of distinct cell types within each area and of different areas within the network is still unknown.

To address these issues, here we recorded the activity of single neurons from a set of interconnected parieto-frontal areas (AIP, F5, F2 and F6) from three macaques trained to perform a visuomotor grasping task in both light and dark conditions. In the grasping-in-the-dark condition, the three-dimensional features of the observed objects had to be temporarily stored in memory in order to shape the hand appropriately and to perform the grasping task correctly.

By applying population decoding analyses and examining the stability of object coding over time in specific subpopulations of neurons within each area, we found that area AIP exhibits a clear and persistent tuning for the object across the different phases of memory-driven trials, area F6 exhibits a more sensory-driven and temporally specific coding, whereas F5 and F2 display intermediate tuning features.

These results suggest that local and long-range neuronal mechanisms cooperate in exploiting the sensory- and memory-driven transformations of object physical properties in the appropriate motor plans for manual actions in primates.