Environmental Orientation as a Prior in Perceptual Depth Assignment: An Information- Theoretic Analysis of Axis Stabilization in Ambiguous 3D Stimuli
Abstract
Perception of three-dimensional structure from two-dimensional images is inherently underdetermined, requiring the visual system to rely on prior constraints to stabilize unique interpretations. Ambiguous wireframe polyhedra, such as the Necker cube, provide minimal stimuli for investigating how non-visual contextual information biases depth assignment. While existing models emphasize image-based heuristics and learned visual statistics, persistent physical regularities of the environment may also shape perceptual priors. In this work, we propose that environmental orientation cues associated with gravity function as contextual priors that bias representational partitioning of ambiguous stimuli. We formalize perceptual interpretation as probabilistic selection among competing depth partitions and quantify partition stability using mutual information and conditional entropy. When visual likelihoods are symmetric, contextual alignment with environmental orientation increases informational coherence and stabilizes particular interpretations. We derive testable predictions involving body orientation, vestibular conflict, and virtual gravity manipulation, distinguishing environmental-frame priors from purely visual heuristics. The model does not posit any direct physical influence of gravity on neural processing nor propose new physical laws. Instead, gravity is treated as a dominant source of environmental statistical structure shaping perceptual inference through experiential priors. These results suggest that perceptual stabilization reflects not only internal computational constraints but also long-term alignment with persistent physical asymmetries of the environment, with implications for perception in altered gravity contexts and immersive virtual environments.
Fields of Study & Subjects
Visual perception and processing mechanisms
Advanced Vision and Imaging
Face Recognition and Perception
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 1 January 2026
- Document type
- Preprint