From retinal size to semantics: Integration of size information into object-representations

Poster Presentation 26.414: Saturday, May 16, 2026, 2:45 – 6:45 pm, Pavilion
Session: Attention: Features, objects

Kelly McEvoy1, Dick Dubbelde2, Sarah Shomstein1; 1The George Washington University, 2Georgetown University

Semantic knowledge of an object’s real-world size influences how attention interacts with objects, yet the mechanism underlying these effects remains unclear. Real-world size is a unique representational dimension because it can be encoded in two dissociable ways: retinal size, reflecting the visual angle projected on the retina, and inferred size, reflecting observers’ knowledge of an object’s true physical dimensions. Whereas retinal size is a low-level sensory measure, inferred size reflects higher-level contextual information. We hypothesize that real-world size shapes both perceptual processing as well as attentional selection. Experiment 1, using rectangles of different retinal size, replicated the classic finding that targets appearing within larger objects are processed less efficiently than those within smaller objects, demonstrating a robust effect of retinal size on perceptual processing. Experiment 2 tested whether this size-based cost also emerges when only inferred size differs. Using line drawings matched in retinal extent, responses were slower for semantically large objects (e.g., door) compared to semantically small ones (e.g., domino), indicating that a size effect originally arising from low-level inputs also manifests when size is defined purely by semantic knowledge. Together, these experiments support the hypothesis that attentional effects elicited by retinal size become integrated into higher-level object representations. Experiment 3 asked whether inferred size further constrains attentional allocation by probing object-based attention. In an adapted two-rectangle paradigm using small- and large-sized real-world object pairs, robust object-based effects were observed. Critically, the magnitude of the same-object advantage was significantly smaller for semantically small than semantically large objects. This pattern indicates greater attentional prioritization for smaller inferred-size objects. Across three experiments, our findings show that real-world size consistently modulates attentional performance, suggesting that size, both retinal and inferred, is embedded within object representations and systematically governs how attention is deployed.

Acknowledgements: NSF BCS 2022572 to SS