Celine Aubuchon¹, Roland W Fleming^1,2,3; ¹Justus Liebig University Giessen, ²Center for Mind, Brain and Behavior, Universities of Marburg, Giessen & TU Darmstadt, Germany, ³Excellence Cluster “The Adaptive Mind”, Universities of Marburg, Giessen and TU Darmstadt, Germany

While image patterns that elicit 3D percepts (e.g., shading, texture) are typically described as “depth cues”, they often do not actually specify depths. Instead, different cues are systematically related to different 3D properties. For example, shading relates primarily to an object’s second-order surface structure, but does not capture how stretched or sheared it is (first-order structure). Conversely, texture gradients vary directly with first-order structure, but are often too coarse to capture subtle curvatures. This means that in typical conditions, when both cues are present, they actually provide complementary information about shape. How are these naturally occurring “conflicts” about depth structure resolved? Put another way: how are the complementary signals integrated? To investigate this, we designed a class of 3D objects that maximize perceptual shape differences between shading and texture. Smooth, thin blobs were first perturbed by shallow second-order corrugations along one-direction and then strongly sheared in the orthogonal direction. This leads to shapes that look quite different depending on whether they are shaded or textured. Observers adjusted 3D probes to report the perceived shape across different surfaces in shading-only, texture-only, and combined-cue conditions. As expected, in the shading conditions observers reliably perceived the corrugations, but saw the object as largely fronto-parallel rather than sheared. Conversely, in the texture conditions, shear was apparent, but the subtle corrugations appeared flatter. Interestingly, when both cues were available, observers’ perceived shapes were not a compromise between the two single-cue estimates, but rather retained the more informative aspects of both cues. Indeed, depth estimates were closer to shading in one direction and closer to texture in the orthogonal direction within the same stimulus. This has important implications for cue-integration models, because it suggests that cues do not necessarily supply “redundant” estimates, but instead may be assigned specialized roles in estimating particular aspects of shape.

Acknowledgements: This study was funded by the European Research Council (project number 101098225 - ERC-2022-AdG “STUFF”)