Cai-Xia Chen¹, Dan-Qing Jiang¹, Xing-Nan Zhao¹, Xin Wang¹, Shi-Ming Tang¹, Cong Yu¹; ¹Peking University

Orientation discrimination of a line target is impaired when flanking lines are added. This crowding effect is at least partially attributable to neuronal surround suppression. The Michael Herzog lab also reported that when the target and flanking lines differ in length, the resultant separate Gestalts reduce crowding, suggesting that surround suppression is modulated by Gestalt organization. We investigated the neuronal responses associated with this de-crowding effect. Using two-photon calcium imaging, we simultaneously recorded responses of large populations of V1 and V4 neurons at various target-flanker distances, with the flanker set to the same or longer length, in awake, fixating macaques. The target was a 0.8° long line presented in the parafovea (1.2-3.4° eccentricity), while the flankers were a pair of parallel lines either the same length as the target or longer (4°). The results revealed: (1) Compared to V1 neuronal responses to the target line alone, surround suppression was significant with the addition of flanking lines. However, with longer flanking lines, less suppression was evident at certain target-flanker distances. (2) Using a SVM to classify the same and longer flanker lengths based on PCA-transformed population neuronal responses to the target line, a decoding accuracy of 70% could be reached across all target-flanker distances with up to 25 PCs. (3) Upon reducing the population neuronal responses to a low-dimensional space using multidimensional scaling (MDS), the distances between different orientations were greater under the longer flanker condition compared to the same-length flanker condition. (4) In contrast, V4 neuronal responses were further suppressed by longer flanking lines. We conclude that Gestalt modulation not only reduces surround suppression but also enhances the separability of target orientation at the population representation level. The opposing effects seen in V4 suggest that the Gestalt modulation effects in V1 may not be driven by downstream feedback mechanisms.

Acknowledgements: STI2030-Major Projects grant (2022ZD0204600)