Perceptual Learning Remains Task Specific with TPE Training

J.-Y. Zhang,L.-J. Cong,C. Yu
DOI: https://doi.org/10.1167/14.10.950
2014-01-01
Journal of Vision
Abstract:Visual perceptual learning is known to be specific to the trained retinal location, orientation/direction, and task. However, in previous studies we have used new "double training" and "training-plus-exposure" (TPE) training protocols to enable complete learning transfer to untrained locations and orientations/directions. Here we examined the task specificity issue by using a TPE protocol to test the hypothesis that perceptual learning may contain a task-independent component, and task specificity may result from an observer's unfamiliarity of the untrained task. Perceptual learning of contrast and orientation discrimination for a foveal Gabor and their mutual transfers were studied. The observers were trained with one task at threshold. Simultaneously in the same session in alternating blocks of trials or in later sessions they also performed and familiarized the other task at suprathreshold (4.5 x thresholds) (The exposure condition). Our results show: (1) Without the exposure condition both orientation and contrast discrimination learning was task specific, replicating known task specificity; (2) When the exposure condition was added, learning showed significant transfer to the other task, but the transfer could be accounted for by the effect of combined pretest and suprathreshold task practice, with no statistically significant evidence for extra benefits from the TPE (training plus exposure) protocol. These results suggest the boundary of our new training protocols in enabling the transfer of specific perceptual learning and support the known conclusion that perceptual learning is task specific, at least for our trained tasks, even with the TPE protocol. The results are consistent with our theory that during perceptual learning the observers learn the rules of performing a specific task. Such rules are applicable to untrained retinal locations and orientations/directions, but not to an untrained new task. Meeting abstract presented at VSS 2014
What problem does this paper attempt to address?