Conjunctive tuning and compensatory cortical distortions support a low-dimensional mechanism of predictive visual remapping

Poster Presentation 36.418: Sunday, May 17, 2026, 2:45 – 6:45 pm, Pavilion
Session: Eye Movements: Models, remapping

Xize Xu1, Sachira Denagamage2, Anirvan Nandy1, Monika Jadi1; 1Yale University, 2Columbia University

Perceptual continuity across saccades depends on pre-saccadic receptive fields (RF) remapping of visual neurons, a process driven by self-motion signals – corollary discharge (CD) – from eye-movement planning areas. Despite prior computational efforts, key gaps remain in our understanding of the underlying mechanisms: (1) What is the role of pre-saccadic suppression (PS)—the temporal reduction in neuronal firing rates during saccade planning—in remapping? (2) Do neurons remap independently or in a coordinated manner consistent with low-dimensional population dynamics? (3) Most importantly, visual areas that perform RF remapping exhibit highly retinotopic organization, characterized by eccentricity-dependent cortical magnification (CM) and contracted representation of polar angles from the visual field. How do these non-uniform geometric characteristics impact the accuracy of remapping? In this study, we present a network mechanism whereby CD inputs guide RF forward remapping in a recurrent population model of 2-D cortical space, represented by neurons conjunctively selective to planned saccade direction and visual field location. We reveal a regulatory role of PS to counteract the adverse effects of saccade direction selectivity during saccade planning. By yoking the responses of neurons through structured recurrent connectivity, our model predicts conserved cell-cell RF relationships across remapping. We show that this prediction and the assumption of saccade direction selectivity are supported by simultaneous population recordings in area V2 from macaques performing a cued saccade task. Cortical transformations such as eccentricity-dependent CM and polar angle contraction systematically distort remapping in our model, which is again confirmed by the empirical data. Finally, we show that remapping distortion due to CM is mitigated by distortion due to polar angle contraction in our model, a phenomenon explained by the counteracting nature of distortion from the visual field to cortical representation by each factor.