Limb position drift results from misalignment of proprioceptive and visual maps

Previous work (Brown et al., 2003a,b) has shown that limb position drifts when individuals make repetitive movements in the absence of visual feedback. The purpose of this study was to examine whether limb position drift might reflect a misalignment in visual and proprioceptive maps by examining the nature of information used to specify new movements from a drifted limb position. In a virtual reality (VR) environment, participants made continuous movements with their dominant right hand between two targets positioned 15 cm apart, paced by a 0.625-Hz metronome. After 5 cycles, cursor feedback of the hand was removed for the next 44 cycles, which induced an average drift in hand position of roughly 5 cm. On the 50th cycle, participants were required to move to one of 6 new targets from the drifted hand position. Kinematic analysis indicated that movement direction was unambiguously determined by the visual input marked by the original start position, or the last-seen hand position. Forward dynamics analysis revealed that current limb configuration was used to inform joint torques to produce this parallel direction. For new movement specification, accurate proprioceptive information about the drifted limb position was used, even though it was apparently not available for detecting drift in the first place. Movement distance varied directly with the extent of limb drift, although the differentiation of visual and proprioceptive control of distance could not be analyzed, as our control conditions were not significantly different for this measure. We suggest that movement drift, in the absence of visual feedback during cyclic repetitive movements, reflects a misalignment between largely accurate visual and proprioceptive maps, rather than a weighted fusion of the two modalities.