On learning two different tasks of same input stimulus

Abstract

Characteristics of perceptual learning, particularly specificity or generalization, can be influenced by the taxonomic relationship between training and testing tasks and the performance level in the training and testing stages. To better illustrate these characteristics, we developed a set of global orientation/motion judgment tasks of identical stimulus, measured subject's performance at two tasks of twenty-five orientation/motion coherence intensities, and trained subjects to report either orientation or motion direction with the other untrained dimension (i.e. motion or orientation) at four different coherence levels. Four groups were trained on a global orientation discrimination task for eight days and the coherence levels of zero, threshold at 63% correct, threshold at 79%, and suprathreshold coherence were assigned to the task-irrelevant feature (i.e. global motion) in each group. Similarly, another four groups trained on a motion direction discrimination task and exposed to four different orientation coherence levels. We found that (1) subjects' thresholds for the trained task improved significantly, mostly around the trained performance level; (2) learning outcomes at the trained dimension were irrespective of the exposed level of untrained dimension; (3) training with exposure to a certain coherence level of untrained dimension can almost fully generalize to the same task but with different levels of untrained dimension; (4) no significant improvement in motion direction task was observed when subjects were trained on orientation discrimination task, no matter to which motion coherence level they were exposed; (5) For motion direction learning groups, the learning effect could significantly but only slightly generalize to orientation tasks, which is also independent of the exposed orientation coherence level. No significant gain in both tasks was found in the control group. Our results suggest that perceptual learning may not result from the strengthened representation at early stage but depends on the involvement of multiple cortical areas and complex neural networks.

Meeting abstract presented at VSS 2017