Dexterous Manipulation: From High-Level Representation to Low-Level Coordination of Digit Forces and Positions

Abstract

The ability to perform fine object and tool manipulation, a hallmark of human dexterity, is not well understood. We have been studying how humans learn anticipatory control of manipulation tasks to characterize the mechanisms underlying the transformation from multiple sources of sensory feedback to the coordination of multiple degrees of freedom of the hand. In our approach, we have removed constraints on digit placement to study how subjects explore and choose relations between digit forces and positions. It was found that the digit positions were characterized by high trial-to-trial variability, thus challenging the extent to which the Central Nervous System (CNS) could have relied on sensorimotor memories built through previous manipulations for anticipatory control of digit forces. Importantly, subjects could adjust digit forces prior to the onset of manipulation to compensate for digit placement variability, thus leading to consistent outcome at the task level. Furthermore, we found that manipulation learned with a set of digits can be transferred to grips involving a different number of digits, despite the significant change in digit placement distribution. These results have led us to propose a theoretical framework based on high-level representation of manipulation tasks can be learned in an effector-independent fashion and transferred to some, but not all that contexts. We discuss these findings in relation to the concept of motor equivalence and sensorimotor integration of grasp kinematics and kinetics.