A Trick Invented in Florence

Some time around 1425, Filippo Brunelleschi stood in the doorway of Florence cathedral holding a small painted panel of the Baptistery opposite. He had drilled a hole through the panel at the vanishing point. You looked through the hole from the back, held a mirror at arm's length in front, and saw the painting reflected. Move the mirror aside and you saw the building itself.

They matched.

That demonstration is the founding event of Western pictorial space, and what makes it so effective is precisely what people forget about it afterwards. It worked because the apparatus fixed one eye at one point and stopped it moving. Perspective is exact for that eye and wrong for every other, which is why the columns at the edges of a wide perspective construction look stretched when you stand where an actual person stands.

Alberti wrote the method down in 1435, in On Painting, with the formulation everything else has been arguing with ever since: the picture is a window through which the subject is seen. Draw a rectangle, decide the height of your eye, put the vanishing point there, and every parallel line running away from you converges on it.

The technical payoff is enormous. Space becomes measurable, figures can be placed at determinate distances and scaled correctly, and a room can be built in the picture that behaves consistently. Piero della Francesca wrote a treatise on the mathematics. You can reconstruct the real dimensions of a painted room from a good perspective picture, because the construction is reversible.

There is also a social payoff, and it is not incidental. Painting in fifteenth-century Italy was a craft, organised in guilds alongside the trades: Florentine painters belonged to the guild that included apothecaries, because that is who sold pigment. A geometric procedure with a written theory behind it gave painters a claim to be practising a liberal art, an intellectual discipline rather than a manual skill. The treatises are making that argument as much as they are teaching the method.

Now the cost.

The system assumes a single, static, monocular viewer. All three assumptions are false about a person in a room, and each is a choice that can be declined.

Chinese landscape handscrolls use a shifting viewpoint. The scroll is unrolled section by section, so there is no single moment at which the whole image is before you; a fixed vanishing point would be answering a question the format does not ask. The convention there is that you travel through the landscape rather than stand outside it, and the multiple viewpoints are how that is achieved.

Byzantine icons frequently use inverse perspective, where lines diverge as they recede, so the space seems to open towards the viewer. This is not incompetence. The makers had access to classical illusionism and its remains were on the walls around them. An icon is not a view through a window. It is a presence addressing you, and a construction organised around the viewer's physical position would make the wrong claim about what is happening.

Egyptian painting shows each part of a body from its most characteristic angle because the image's job is to specify what is there, not to record how it looked from one spot on a particular afternoon.

So perspective is one answer to the question of what an image is for, not the correct solution to a problem everyone else was failing. The habit of reading it as progress is a nineteenth-century idea about art history, and it produces a flattened reading of nearly everything outside a three-hundred-year European window.

The system's own limit is best seen where a painter tests it. In Holbein's Ambassadors of 1533, two men stand among the instruments of learning, rendered with immaculate perspectival control, and across the bottom of the picture lies a long grey smear. Stand to the right, close to the panel, and the smear resolves into a skull.

You cannot see both at once. From the position the painting was built for, death is an unreadable distortion; from the position where death is legible, the ambassadors' world is skewed into nonsense. The viewer's place, which perspective normally treats as a silent assumption, has been made the subject.