A shadow that is too dark, a highlight that is too bright, a hallway that narrows too fast: in many celebrated paintings, light, color, and perspective are quietly wrong. Yet the image feels not broken but hyper-real, as if the scene were turned up a notch beyond everyday vision.
Vision science offers a map of this paradox. The eye delivers noisy data to the cortex, which relies on predictive coding and Bayesian inference to guess the most probable world. Painters exploit that guesswork. By exaggerating luminance contrast and chromatic contrast beyond what optics would allow, they feed the brain signals that align with its internal model of how edges, depth, and materials should behave.
Linear perspective is bent in similar fashion. Many works splice several vantage points into one frame, breaking geometric projection while preserving optic flow cues that the visual system uses to compute spatial layout. Local distortions are tolerated because global coherence stays intact, so the scene is tagged as stable. Color is also rigged through color constancy: surfaces are painted with physically impossible hues, but arranged so relative relationships still match the brain’s schema for skin, stone, or sky.
The result is less a copy of the external world than a user interface for the visual cortex, tuned to its shortcuts, biases, and expectations rather than to physical accuracy.
