Multimodal Encoding of Goal-Directed Actions in Monkey Ventral Premotor Grasping Neurons

Visuo-motor neurons of the ventral premotor area F5 encode "pragmatic" representations of object in terms of the potential motor acts (e.g., precision grip) afforded by it. Likewise, objects with identical pragmatic features (e.g., small spheres) but different behavioral value (e.g., edible or inedible) convey different "semantic" information and thus afford different goal-directed behaviors (e.g., grasp-to-eat or grasp-to-place). However, whether F5 neurons can extract distinct behavioral affordances from objects with similar pragmatic features is unknown. We recorded 134 F5 visuo-motor neurons in 2 macaques during a contextually cued go/no-go task in which the monkey grasped, or refrained from grasping, a previously presented edible or inedible target to eat it or placing it, respectively. Sixty-nine visuo-motor neurons showed motor selectivity for the target (35 food and 34 object), and about half of them (N = 35) exhibited congruent visual preference. Interestingly, when the monkey grasped in complete darkness and could identify the target only based on haptic feedback, visuo-motor neurons lost their precontact selectivity, but most of them (80%) showed it again 60 ms after hand-target contact. These findings suggest that F5 neurons possess a multimodal access to semantic information on objects, which are transformed into motor representations of the potential goal-directed actions afforded by them.