Generalization in perceptual learning across stimuli and tasks in varied adaptation levels.

Introduction: Perceptual learning has been established as a method for improved visual perception and for the study of visual processes. While early studies, using the backward-masked texture discrimination task (TDT) focused on the location specificity effect and on the lack of transfer of the learned task, recent TDT studies used conditions that considered to eliminate sensory adaptation (‘dummy paradigm’) showing the transfer. It was suggested that inhibitory processes triggered by adaptation and spatiotemporal context (masking) reduce learning efficiency. The reduced efficiency suggests that TDT learning is limited by non-texture tasks such as backward masking (BM). Here, we address this issue using the TDT method with different adaptation levels (standard and dummy paradigms), and with pre/post-tests of BM. Methods: The texture stimulus training was based on the paradigm used by Karni and Sagi (1991). We compared the backward masking (Gabor target at fixation, masks 2 Gabor flankers), of the subject before and after the TDT training. The stimulus onset asynchrony (SOA) was 80,120, 160, 200, and 240 milliseconds (ms). Results: Standard and dummy TDT learning showed typical threshold improvements. BM pretests before TDT training did not affect initial TDT thresholds in all conditions. Importantly, standard TDT training (high adaptation) reduced effective BM duration, while reduced adaptation TDT training had no significant effect on BM duration (p=0.003). Conclusions: Our results suggest that learning generalization across experimental conditions and tasks depends on shared visual processes, here dominated by inhibitory effects involved in adaptation and masking. Interestingly, increased adaptation, due to increased stimulus duration or stimulus consistency, enabled transfer learning across tasks. The reduced BM effect after TDT training suggests decreased task-independent inhibitory processes due to training.