Sensorimotor Learning

Sensorimotor learning

We conducted a series of studies to understand inter-related phenomena associated with sensorimotor learning of dexterous manipulation, i.e., transfer and interference. The key questions we focused on were: Can learned dexterous manipulation be transferred to a different context, e.g., object orientation or part of an object to be grasped? To address this question, we asked subjects to learn to exert a compensatory torque using a precision grip in an anticipatory fashion, i.e., at object lift onset. After learning the task, subjects were asked to rotate the object 180° to assess the extent to which they could transfer the learned relation between digit forces and position. We found that subjects failed to transfer the learned manipulation on the first object rotation. However, with additional object rotations subjects learned to use the appropriate fingertip position, but not force distribution.1 This suggests separate sensorimotor mechanisms for controlling digit position and forces.

Figure 5. Left: Experimental protocol and sequence of grasp contexts (left or right handle of the U-shaped object). Right: Compensatory torque (Tcom) and peak object roll across trials from each grasp context. From: Fu and Santello (2012).

Further studies addressed the extent to which dexterous manipulation learned in a given context (left or right handle of a U-shaped object) can be transferred in a novel dynamic scenario (opposite handle of a U-shaped object).2,3 We found a significant interference of the previously-experience context when subjects grasped the opposite handle or when the external torque changed direction after changing the object orientation (Fig. 5). We interpreted these finding as evidence that participants did not learn the dynamic of the object per se, but rather the task.


For more information on these projects contact: Marco Santello