ADAPTIVE INTERNAL MODEL OF INTRINSIC COORDINATES
TRANSFORMATION DURING LEARNING OF A REACHING TASK.
H. Imamizu*, Y. Uno and M. Kawato
ATR Human Information Processing Research Labs., Kyoto, Japan
Recent computational studies have proposed that the central nervous system acquires internal models of coordinates transformation between task-oriented extrinsic space and intrinsic space such as joint angles. To investigate acquisition of the internal model, we virtually minified (1/2) the elbow angle and magnified (5/4) the shoulder angle of human subjects while they were aiming at targets. A position marker was attached to the subject's hand and its current altered position was displayed as a cursor on a CRT screen. This linear transformation in joint angles (intrinsic coordinates) corresponds to a nonlinear one between the hand plane and the screen (extrinsic coordinates). We investigated whether the subjects learn this transformation as the former or the latter one. The aiming error when they learned the transformation of joint angles of one arm using the same arm (consistent condition) was compared to that when they learned the same transformation using the opposite arm (inconsistent condition). The error in the inconsistent condition was significantly larger (p <.002) than that in the consistent condition. Furthermore, the error in the consistent condition became small after they learned the transformation using the opposite arm in the same condition compared with that when they learned it for the first time (p <.0001). That is, intermanual transfer of the learning effect was found in the consistent condition but not in the inconsistent condition. Results suggest that the subjects learned the transformation as a linear one in intrinsic coordinates in the consistent condition and that the central nervous system adaptively represented the transformation including intrinsic coordinates in the control of arm movements.