Linear perspective
A Socratic walk-through of linear perspective — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #How can lines on a flat surface create a convincing impression of depth?
A painting of a corridor is as flat as a painting of a flag, yet one of them recedes. Nothing physical distinguishes them; both are pigment on a plane. So the depth is not in the surface but in the arrangement — which raises a question worth sitting with. What does your eye receive from a real corridor that a flat surface could reproduce exactly?
Reasoning it through
REASONING #Consider what reaches the eye at all. Light travels in straight lines, so the eye does not receive the corridor; it receives a bundle of rays converging on a point. Now the decisive question. If you slid a sheet of glass into that bundle and marked where each ray pierced it, what would the eye receive afterwards? The same bundle, because each ray still arrives from the same direction. That is the whole trick, and it is geometric rather than artistic.
The rule is central projection: every scene point joined to one eye point, the image being where those joins meet the picture plane. Two consequences follow, both provable with similar triangles.
First, size. A thing twice as far away subtends half the angle, so its mark is half as tall — smaller by a factor, in inverse proportion to distance, not by a fixed amount. Notice what that does to a row of equally spaced things, paving stones or fence posts. The first gap is halved, the next much less than halved, and the marks crowd toward a limit rather than marching evenly toward it. Would you have guessed that by eye? Drawing freehand, most people space receding stones far too evenly, and the floor tilts up instead of going back.
Second, direction. Take a set of parallel lines running away from you. As points on them get further off, their images crowd toward one place on the plane — where a line drawn from the eye parallel to that set pierces it. Every line of the set aims there. So parallel lines meet in the picture, and where they meet is fixed by their direction, not by the objects: change direction and the point moves, while lines parallel to the picture plane never pierce it at all, which is why the verticals of a building drawn straight-on stay parallel on the page. All the horizontal ground directions give points lying on one line at eye height — so the horizon is not the edge of the earth but a statement about where the painter's eye was.
The analogy
THE ANALOGY #Alberti described a picture as an open window. Take him literally: hold a pane of glass still, close one eye, keep your head fixed, and trace what you see onto the glass. The finished tracing sends your eye the same directions the world did. You have not copied objects; you have recorded where their rays crossed a surface.
the tracing is exact only from the spot your head occupied, and it can never carry what two eyes and small movements would have told you — the disagreement between the views, and the sliding of near things against far. A real window supplies those; the traced pane quietly supplies evidence that it is flat.
Clarifying the model
THE MODEL #Which points at the misconception worth correcting. Perspective does not make a picture look like reality; it makes the picture send the eye the directions reality would have sent, and only from the intended viewpoint. Almost nobody stands there, yet pictures survive it well — viewers appear to register the surface's own slant and compensate, a robustness still studied and not fully explained. That tolerance is a fact about viewers, not a property of the geometry.
Two further clarifications. Perspective is one depth cue among several and often not the strongest: overlap, texture growing finer with distance, shading, and the haze that pales distant hills all work in a flat picture too, and a painting can be spatially convincing with no converging lines at all. And the geometry was invented, not discovered by people who had failed to see — demonstrated by Brunelleschi in Florence in the early fifteenth century, set out as a construction by Alberti in 1435. Traditions that depicted space by other conventions were choosing differently about what a picture is for.
A picture of it
THE PICTURE #How to readEach bar is the same 2-metre figure as it would be marked on a pane held one metre from the eye, at the distance labelled beneath, and the line traces the same values as a curve; the arithmetic is 2 metres divided by the distance. Watch the rate of shrinking rather than the size: the step from 2 m to 4 m costs half the height, while the equally large step from 16 m to 20 m costs barely two centimetres. That flattening is why receding things must be drawn crowding together, and why evenly spaced marks read as a floor tipping upward instead of going back.
What became clearer
WHAT CLEARED #Depth on a flat surface is not an illusion painted into the marks. It follows from putting each mark where a ray from the scene would have crossed a plane on its way to a single eye — and from that one rule everything else falls out: sizes shrinking in inverse proportion to distance, parallel lines converging on a point fixed by their direction, and a horizon that records the height of the eye that looked.
Where to go next
ONWARD #- Why photographs taken with very wide or very long lenses look "wrong" to us, though the geometry is exact.
- How anamorphic images push the single-viewpoint rule to its extreme.
- What curvilinear and multi-point schemes try to fix about straight-line perspective at wide angles.
Key terms
TERMS #| Term | What it means |
|---|---|
| Central projection | mapping each scene point to where the line joining it to a single viewpoint meets the picture plane. |
| Vanishing point | where the line drawn through the eye parallel to a set of parallel lines pierces the picture plane, and so where their images converge. |
| Horizon line | the line holding the vanishing points of all horizontal ground directions, at the height of the viewer's eye. |
Every term the collection defines is gathered in the glossary.