Item

Abstract

The visuomotor system recalibrates when visual and motor maps are in conflict, bringing the maps back into correspondence. For recalibration to occur, a conflict has to be detected. Ernst and Endreß (VSS '04) showed that the rate of recalibration in a one-dimensional visually guided pointing task depends on the uncertainty of the feedback: faster recalibration with less uncertainty. In the present work, we examined two-dimensional recalibration and how the form of visual feedback affects it. Subjects pointed with an unseen hand to a brief visual target. Visual feedback was given indicating where the point landed. We introduced a constant conflict between pointing and feedback location and examined the changes in pointing as the subject adapted. We asked whether differential vertical and horizontal uncertainty in the visual feedback affects recalibration rate differentially, or whether rate is determined by the total uncertainty. We also varied feedback uncertainty in two ways. (1) We blurred the visual feedback, thereby reducing its localizability; in this condition, uncertainty could be determined on-line from one feedback stimulus. (2) We introduced random trial-by-trial perturbations in the feedback; in this condition, uncertainty had to be learned over time. In both cases, the distributions determining the vertical and horizontal uncertainties were 2D Gaussians. Adaptation profiles (changes over time in the point location relative to the visual feedback) changed only in response to changes in localizability. Recalibration was slowest in the direction of greatest uncertainty when uncertainty was due to blur, but rate was unaffected by trial-by-trial variation. This means that subjects do not estimate uncertainty over time in order to adjust reaching. Rather, they adjust trial by trial based mostly on feedback from the previous trial.