Comparing limb proprioception and oculomotor signals during hand-guided saccades.

We previously showed that saccades tend to overshoot briefly flashed targets that were manually displaced in the dark (Ren et al. 2006). However it was not clear if the overshoot originated from a sensory error in measuring hand displacement or from a premotor error in saccade programming, because gaze and hand position started at the same central position. Here, we tested between these hypotheses by dissociating the initial eye and hand position. Five hand/target positions (center, far, near, right, left) on a frontally-placed horizontal surface were used in four paradigms: Center or Peripheral Eye-hand Association (CA or PA, both gaze and right hand started from the center or a same peripheral location) and Hand or Eye Dissociation (HD or ED, hand or gaze started from one of three non-target peripheral locations). Subjects never received any visual feedback about the final target location and the subjects' hand displacement. In the CA paradigm, subjects showed the same overshoot that we showed previously. However, changing both initial eye and hand positions relative to the final target (PA) affected the pattern, significantly altering the directions of overshoots. Changing only the initial position of hand (HD) did not have this effect, whereas changing only initial eye position (ED) had the same effect as the PA condition (CA approximately HD, PA approximately ED). Furthermore, multiple regression analysis showed that the direction of the ideal saccade contributed significantly to the endpoint direction error, not the direction of the hand path. These results suggest that these errors do not primarily arise from misestimates of the hand trajectory, but rather from a process of comparing the initial eye position and the limb proprioceptive signal during saccade programming.