Abstract
Random dot kinematograms (RDKs) have been used to study motion perception for decades, but almost all studies involve a binary choice (e.g., leftward vs. rightward). By allowing the direction to vary over 360° with various coherence levels, and by asking the participant to report the exact direction of motion in continuous 360° space, we were able to observe three interesting phenomena that cannot be observed in the traditional discrete alternative choice task. First, we found that participants often have the illusion that the motion is in the opposite of the true direction (180° errors). Second, the perceived direction of motion is biased away from the cardinal axes (which may serve as implicit category boundaries). Third, observers make two distinct kinds of errors: normally distributed errors around the true motion direction, and uniformly distributed errors over the all possible motion directions. This suggests that behavioral performance reflects a mixture of trials in which the motion was perceived with some degree of precision and trials in which the true stimulus direction was not detected at all. By applying a mixture model, we were able to estimate the precision on trials where the motion was detected and the probability that the motion was not detected. Both of these parameters were systematically influenced by the coherence level and the stimulus duration. Together, these findings provide new insights into the perception of motion and into perceptual decision-making in a context where the decision space is continuous rather than discrete.
Meeting abstract presented at VSS 2018