Location and direction specificity in motion direction learning associated with a single-level method of constant stimuli.

Recent studies reported significantly less location specificity in motion direction learning than in previous classical studies. The latter performed training with the method of constant stimuli containing a single level of direction difference. In contrast the former used staircase methods that varied the direction difference trial by trial. We suspect that extensive practice with a single direction difference could allow an observer to use some subtle local cues for direction discrimination. Such local cues may be unavailable at a new stimulus location, leading to higher location specificity. To test this hypothesis, we jittered slightly the directions of a stimulus pair by the same amount while keeping the direction difference constant, so as to disturb the potential local cues. We observed significantly more transfer of learning to untrained locations. The local cue effects may also explain the recent controversies regarding the finding that foveal motion direction learning becomes significantly more transferrable to a new direction with TPE (training-plus-exposure) training. One specific study by Zili Liu and collaborators that challenges this finding also used a single-level direction difference for training. We first replicated their results. But we found that if the directions of the stimulus pair were again jittered while the direction difference was kept constant, motion direction learning transferred significantly more to an orthogonal direction with TPE training. Our results thus demonstrate the importance of using appropriate psychophysical methods in training to reduce local-cue related specificity in perceptual learning.