Lisa Schwetlick^1,2, Peter Neri²; ¹EPFL, ²ENS

The ability to discriminate visual features is profoundly influenced by the surrounding context. Understanding center-surround dynamics is essential for uncovering the mechanisms of contextual interactions that play a role in natural scene processing. Previous research has highlighted complex dependencies of orientation tuning on the spatial and temporal relationships between a central probe and its surround. Here, we extend this work by investigating how orientation noise and contextual configurations influence orientation discrimination in a 2AFC task. Participants identified which of two patches contained a target orientation while the surround elements varied in shape (full ring, four diagonal petals, or two opposing petals) and grating content (aligned with target orientation, orthogonal to target orientation, or plaid combining both). Both patches were independently corrupted by orientation noise to support reverse correlation analysis. In the absence of a surround, noise aligned with target orientation increased the probability of classifying the corresponding patch as containing the target, while anti-target noise orthogonal to the target reduced it. In the presence of a plaid surround, these noise effects remained similar, but were attenuated. When the surrounds contained a target-aligned grating, the effect of target-aligned noise was selectively reduced, while surrounds with orthogonal-to-target gratings reduced the impact of orthogonal-to-target noise. While surround effects were mainly driven by surround content (grating orientation), we also observed subtle effects of surround shape (petal alignment with the target). Moreover, these effects become evident only when the surround is presented at the same time as- or after the target, but not when the surround is shown before the target. These results reveal nuanced spatiotemporal contextual effects that dynamically modulate local orientation tuning, and provide critical constraints for developing models of contextual interactions in vision.