Vector coding in slow goal-directed arm movements.

It has been found that the estimate of relative target direction is consistently biased. Relative target direction refers to the direction in which a target is located relative to another location in space (e.g., a starting position in the case of goal-directed movements). In this study, we have tested two models that could underlie this biased estimate. The first proposed model is based on a distorted internal representation of locations (i.e., we perceive a target at the "wrong" location). We call this the distorted location model. The second model is based on the idea that the derivation of target direction from spatial information about starting and target position is biased. We call this the biased direction model. These two models lead to different predictions of the deviations that occur when the distance between the starting position and the target position is increased. Since we know from previous studies that the initial direction of slow arm movements reflects the target direction estimate, we tested the two models by analyzing the initial direction of slow arm movements. The results show that the biased direction model can account for the biases we find in the target direction estimate for various target distances, whereas the distorted location model cannot. In two additional experiments, we explored this model further. The results show that the biases depend only on the orientation of the line through starting position and target position relative to the plane through longitudinal head or body axis and starting position. We conclude that the initial part of (slow) goal-directed arm movements is planned on the basis of a (biased) target direction estimate and not on the basis of a wrong internal representation of target location. This supports the hypothesis that we code displacements of our limbs in space as a vector.