Perceptual learning transfer over space and orientation

Purpose. Early perceptual learning, the gradual performance improvement with training found for many simple visual tasks, was found to be specific to the training stimulus location in the visual field, and also to the orientations of the stimulus elements. We now ask whether there are cases where training effects transfer across space and orientation, and, if so, are they related. Methods. The behavioral task was to detect whether there was an oddly oriented element in a 7×7 array of homogeneous distractors. The limiting factor was the Stimulus Onset Asynchrony (SOA) between test stimulus and subsequent mask. We used a range of SOAs (16-180ms) and the target element could appear in any position, or in one of two positions (on two sides of fixation), for different subject groups, respectively. Following learning, all subjects were tested with the pop-out target appearing in all locations, first with the original orientations, and subsequently with swapped target and distractor orientations. Results. The improvement for subjects trained with the target in two locations did not transfer to more distal positions. However, there was significant improvement for the locations between and around the trained positions. Surprisingly, we found that performance was as good with swapped orientations as with the original orientations. In contrast, swapping orientations had a large detrimental effect on subjects trained with the target everywhere. Thus, the group trained two locations performed much better with swapped orientations than did the group trained on all locations. Conclusions. Subjects trained on two locations were able to transfer their learning effects to new locations and to swapped orientations. Thus, sparse training with respect to one dimension may induce generalization with respect to another.