Asymmetric Transfer of Task Dependent Perceptual Learning in Visual Motion Processing

The effects of perceptual learning (PL) on the sensory representation are not fully understood, especially for higher–level visual mechanisms more directly relevant to behavior. The objective of this research is to elucidate the mechanisms that mediate task dependent learning by determining where and how task dependent learning occurs in the later stages of visual motion processing. Eighteen subjects were trained to perform a dual–2TAFC visual discrimination task in which they were required to simultaneously detect changes in the direction of moving dots (task–1) and the proportion of red dots (task–2) shown in two stimulus apertures presented in either the left or right visual field. Subjects trained on the direction discrimination task for one of two types of motion, global radial motions (expansion and contraction) presented across stimulus apertures (global task), or an equivalent (local) motion stimulus formed by rotating the direction of motion in one aperture by 180°. In task–1 subjects were required to indicate whether the directions of motion in the second stimulus interval were rotated clockwise or counter–clockwise relative to the first stimulus interval. In task–2, designed to control for the spatial allocation of attention, subjects were required to indicate which stimulus interval contained a larger proportion of red dots across stimulus apertures. Sixteen of the eighteen subjects showed significant improvement on the trained tasks across sessions (p\u3c0.05). In subjects trained with radial motions, performance improvements transferred to the radial motions presented in the untrained visual field, and the equivalent local motion stimuli and untrained circular motions presented in the trained visual field. For subjects trained with local motion stimuli, learning was restricted to the trained local motion directions and their global motion equivalents presented in the trained visual field. These results suggest that perceptual learning of global and local motions is not symmetric, differentially impacting processing across multiple stages of visual processing whose activities are correlated. This pattern of learning is not fully coherent with a reverse hierarchy theory or bottom–up model of learning, suggesting instead a mechanism whereby learning occurs at the stage of visual processing that is most discriminative for the given task