Adaptation of reaching movement toward a target to a novel environment can be generalized into the movement aiming toward other targets (Ghahramani et al., 1996). Previous works investigating the generalization pattern of the adaptation to a visual rotation imposed on the cursor have shown that the learning effects decay symmetrically as the movement direction get more distant from the movement direction for the training (Krakauer et al., 2000; Izawa et al., 2012). Notably, an identical training effect could be attained by different training schedules: for example, both abrupt application of 30 deg visual rotation and the gradually increasing visual rotations up to 30 deg should lead to the identical training outcome. However, it remains unknown how these differences in training schedule influences the resultant generalization pattern of motor learning.
In order to examine this problem, subject trained to perform horizontal reaching movement by a novel training schedule. In this schedule, the target location was gradually shifted from the straight-ahead position in 30 deg counter-clockwise (CCW) direction (0.5 degrees per trial), and the same amount of clockwise (CW) visual rotation was also imposed to the cursor to maintain the hand movement to be straight-ahead direction.
We examined how the attained adaptation spatially generalized by measuring the hand movement direction at the peak hand velocity when subjects aimed to different targets. As compared to the generalization pattern reported previously (Krakauer et al., 2000), the training effect was generalized more broadly. Furthermore, we observed more asymmetric generalization pattern in which the generalization was greater when aiming to the target located CCW from the target. Theoretically, the generalization pattern is assumed to be determined by the shape of the motor primitives encoding the movement direction (Thoroughman & Shadmehr, 2000). However, by this framework, it was quite hard to explain the dependence of generalization pattern on the training schedule, indicating that we need to reconsider how the movement direction and/or plan is encoded in the motor primitives.