Abstract
The problem of reference frame coding for visual-to-motor transformations has been widely investigated using a variety of stationary stimuli and visuomotor tasks. Here we investigate the reference frame(s) for motion direction perception in human subjects. We recorded behavioral responses in 4 subjects while they performed a direction ‘perception’ + saccade (experiment 1) and a direction discrimination + key-press-2AFC (experiment 2) tasks. In experiment 1 a random dot pattern (RDP) moving in one of 10 possible directions (from 0[deg] in steps of 10[deg] to 90[deg]) was presented to the subjects during 400ms. After stimulus offset, subjects rotated their heads 30[deg] to the left around the anterior posterior axis and made a saccade in the perceived direction. When comparing to a control condition, in which the head remained stationary, we found that saccade trajectories, mean saccade landing positions and saccade directional errors were not affected by head rotations. In experiment 2 subjects compared the direction of two RDPs presented during 400ms each and separated by a 2000ms time interval. We obtained psychometric curves in two conditions, a) when subjects kept the head straight during the 2000ms interval, and b) when subjects rotated the head 300 to the left during that interval. We found that on average the curves point of subjective equality in the latter condition was slightly displaced in the direction of the head tilt (from −1.78[deg] with the head straight to -2.76[deg] with the head rotated). However, this displacement was considerably smaller than the one predicted by a head/eye-centered model of motion direction perception and much closer to the predictions of a body/space centered model. These results argue against the hypothesis that visual motion is encoded in a head/eye-centered frame of reference. On the other hand, they suggest that motion direction is encoded in a body/space-centered reference frame.