Danny Lim

Presentation Title

 Geometric Context Distorts Visual Spatial Working Memory 

Abstract

Visual working memory is often assumed to rely on retinotopically organized representations, yet accumulating evidence suggests that remembered locations are influenced by contextual structure. Here, we test whether simple geometric boundaries systematically distort spatial working memory. Participants performed a memory-guided localization task in which targets were presented within apertures of different shapes (circle, square, tall rectangle, wide rectangle). Across both saccadic and mouse-based responses, we observed geometry-dependent distortions in recalled locations. Errors were not uniform: rectangular apertures produced anisotropic biases aligned with their principal axes, while circular and square apertures yielded comparatively isotropic patterns. Critically, these distortions emerged in directional components of error, revealing structured deviations that are not captured by aggregate measures such as Euclidean distance. These findings suggest that spatial working memory is not purely retinotopic, but is instead shaped by the geometry of the surrounding context. We propose that boundary structure constrains internal spatial representations, potentially reflecting interactions between retinotopic coding and higher-level, context-sensitive reference frames.