6 ideas
Artists never start from raw visual sensation; they start from an inherited schema, a conventional formula for a tree, face or horse, and then correct it against what they observe. Naturalistic art advances through this cycle of schema and correction, so every realistic image is a revised convention and not a direct copy of the world.
Dürer drew a rhinoceros he had never seen, working from a sketch and written report, and gave it armor-like plates and scales that the animal does not have. His image was reproduced as a model in natural-history illustrations for a very long time afterward. Gombrich uses it as one vivid case of a familiar formula filling in what the available information did not supply.
Nobody sees neutrally. Perception is a process of testing hypotheses against expectations, and what we already know and expect shapes what we notice. Painting what you truly see is therefore something learned through convention and training, not a return to untutored vision.
A picture only works when the viewer completes it, projecting meaning onto ambiguous marks and filling gaps with their own expectations and knowledge. Impressionist brushwork and suggestive sketches depend on this projection. The image is realized in the viewer's mind, not on the canvas.
In ambiguous figures like the duck-rabbit, the eye flips between interpretations but cannot hold two at the same moment. The same holds for a painting: we can attend to the flat canvas or the illusion of depth, not to both at once. Illusion therefore depends on the viewer choosing one reading over the other.
Gombrich traces the development of naturalistic representation, from Egyptian conventions through Greek foreshortening and Renaissance perspective to Constable's skies, as a series of schemata revised when they failed to meet new demands. He ties these shifts to changes in what images were meant to do, such as the Greek turn toward narrative, not to observation alone.