A Theoretical Framework for Visual Weight: Contrast, Size, and Edge Sharpness as Candidate Determinants of Perceived Dominance in Static Images

A composition feels resolved when one element clearly leads and unstable when several compete without hierarchy, yet the stimulus properties that produce this experience remain poorly formalized. This preprint presents a theoretical framework for that problem. The framework proposes that perceived compositional dominance depends on three candidate variables: contrast relative to a perceptually defined ground (the "canvas"), fractional area, and edge transition sharpness ("hardness"). These variables are combined multiplicatively and embedded within a competitive architecture based on divisive normalization. A provisional chromatic extension accounts for interactions between color and luminance. The framework operates at the level of segmented image elements measured against a perceptual ground — an architecture empirically motivated by findings that pixel-level models fail to predict compositional preferences. A patent application covering the computational implementation has been filed. The theoretical framework, experimental predictions, and all academic content remain freely available for research and educational purposes. The manuscript formalizes a concept long discussed qualitatively in art and design practice and derives seven falsifiable predictions, two preregistered prior to data collection. Related preregistration, stimuli, and materials are available at https://osf.io/dgauz and https://osf.io/kmzgj. Version 3 Update: Revised prior to data collection to add patent disclosure, cite empirical support for element-level architecture from McManus et al. (2011), describe a natural-image implementation direction, and refine language throughout. A full changelog is included in the manuscript. Version 2 Update: Revised prior to data collection to update hardness magnitude estimates, clarify structural assumptions, and better distinguish the framework’s region-level contribution from spatial-frequency approaches.