Internal models underlying fingertip force control during object manipulation in humans

Since sensory information pertaining to an object's weight and weight distribution is not available until after an object has been grasped and lifted from its support surface, skillful manipulation requires predictive scaling of fingertip forces. The predictive control is based on internal models of the object formed during prior manipulatory experience. The present study investigates the extent to which such predictive control achieved with practice of one hand can be generalized to the contralateral hand. Subjects grasped and lifted an object instrumented with force transducers. The object's center of mass (CoM) was displaced 2 cm laterally from the object's center, requiring asymmetric partitioning of tangential force development to prevent tilting. They first performed five lifts using their right hand with the CoM located on a given side with a given weight. Then the subjects translated the object the left side of the body and performed one lift with the left hand. The procedure was repeated five times with the CoM located on each side of the object and with a total object weight of 400 g and 800 g. On the last pre-translation lift, the rate of tangential force development was appropriately higher in the digit opposing the CoM (p<0.05). In contrast, following translation to the left hand, the rate of tangential force development was nearly equal in each digit (p>0.05). Nevertheless the force rates were higher in both digits for lifts with the heavier (800 g) object (p<0.05). The results suggest there is a dichotomy between the generalizability of predictive control based on the object's weight and CoM. We propose that the control may be hierarchical in nature, with parameters that can be globalized representing higher level control and those that cannot (e.g., individuated digit control) representing a lower level control.