Abstract
It has been found that sensitivity to a visual feature is enhanced when the feature is repeatedly paired with reward (Seitz, Kim & Watanabe, 2009, Neuron). We call this type of learning reward-driven exposure-based perceptual learning (REPL). In a previous study (Kim, Seitz, Watanabe, 2008, VSS), we presented three different orientations (60 deg separated from each other) which were followed by reward at the probabilities of 80% (positive contingency), 50% (zero-contingency) and 20% (negative contingency), respectively. We found significant performance improvement for both the positive-contingency orientation and zero-contingency orientation, but no significant improvement for the negative-contingency orientation. Given that PL occurs not only as a result of exposure (Watanabe, Sasaki and Nanez, 2001) but also of task-involvements (Fahle & Poggio, 2002), a question arises as to whether reward-driven task-involvement PL (RTPL) occurs in the same way as REPL. To address this question, in the present study, we trained a new group of four subjects with an operant conditioning procedure in which subjects performed an orientation discrimination task, and the reward was given only when the subject answered correctly. To compare these results with REPL, we conducted sensitivity tests before and after operant training. After training, we found significant performance improvement only for the positive contingency orientation. These results suggest that the mechanisms underlying REPL and RTPL are different. One possible model is that when subjects were trained with the RTPL procedure, the learning of 50% reward probability was inhibited by an attentional signal, whereas this inhibition did not occur when they were trained with the REPL procedure.
This research was supported by NIH-NEI (R21 EY018925, R01 EY015980-04A2, R01 EY019466) and NSF-CELEST (BCS-PR04-137).