Yusuke Nakashima¹, Akane Kinumaki¹, Jessica Ding¹, Angie Boza¹, Luciana Chiu¹, Eileen Y Jang¹, Yuka Sasaki¹, Takeo Watanabe¹; ¹Brown University

Specificity to the trained retinal location is a core characteristic of visual perceptual learning (VPL), yet the neural mechanisms underlying location specificity remain unclear. We previously showed that location specificity was eliminated when training on an orientation discrimination task was conducted with backward masking, whereas specificity remained with no masking and with forward masking. Backward masking is thought to disrupt later processing phases, possibly including recurrent processing, whereas forward masking disrupts early phases where feedforward processing may predominate. If these differential roles are valid, the loss of specificity under backward masking suggests that recurrent processing may contribute to location specificity, although direct evidence remains limited. Here, we tested these roles by examining which frequency bands of neural oscillations are affected by backward and forward masking. Neurophysiological studies suggest that theta and gamma neural oscillations carry feedforward signals, whereas alpha and beta bands carry recurrent signals. Thus, if backward and forward masking affect decoding accuracy in recurrent-related (alpha or beta) and feedforward-related (theta or gamma) frequency bands, respectively, such a pattern would support the two-roles model and clarify the role of recurrent processing in location specificity. Participants performed an orientation discrimination task under no masking, backward masking, and forward masking. Decoding accuracy for orientation classification (clockwise vs. counterclockwise) was calculated for theta, alpha, beta, and gamma bands at each timepoint. In the no masking condition, decoding accuracy increased in theta and alpha bands. Backward masking reduced alpha accuracy while sparing theta, whereas forward masking reduced theta and alpha accuracy but left beta accuracy intact. These results suggest that backward masking disrupts recurrent processing reflected in alpha, whereas forward masking disrupts feedforward processing reflected in theta. Together with our previous findings, these results indicate that recurrent processing during training plays an important role in location specificity in VPL.

Acknowledgements: NIH R01EY019466, R01EY027841, R01EY031705, NSF-BSF BCS2241417