Abstract
In perceptual learning, perception improves with practice. Perceptual learning is mainly investigated with retinotopic paradigms. Here, we show evidence for perceptual learning within a non-retinotopic frame of reference. During the training phase, we presented three disks. In the center disk, dots moved upwards with a slight tilt to the left or right. Observers indicated the tilt. Performance improved significantly by 33% on average. Before and after training, we determined performance for various "moving conditions". First, three empty disks were presented at the same location as in the training condition. Then, there was an ISI of 100ms and the disks shifted one disk position to the right (i.e., a Ternus-Pikler display), creating the impression of apparent motion. The same motion as in the training condition was presented either in the left disk, which retinotopically overlapped with the training disk, or in the center disk. The center disk did not overlap with the training disk in retinotopic coordinates but corresponded to it within the object-centered reference frame induced by the Ternus-Pikler motion (the center disk of the first frame maps on the center disk of the second frame). We found that learning transferred only in the latter, non-retinotopic condition. Our results indicate a non-retinotopic, object-centered component to visual perceptual learning. We propose that non-retinotopic learning occurs because both objects and our eyes move constantly leading to retinotopically varying representations. Hence, perceptual learning may best be achieved after spatial invariance is reached.
Meeting abstract presented at VSS 2013