The Verification Venue · a thing every photographer gets told wrong
The Compression That Isn't the Lens
Everyone knows a telephoto lens compresses distance and a wide-angle lens distorts faces. Both effects are real — and neither is caused by the glass. Perspective, the relative size of near and far, is set by one thing only: how far the camera stands from the subject. Drag the focal length below and watch the near/far size ratio refuse to move.
Here is a scene with two identical people — a near one and a far one, standing 20 metres apart in front of a gridded wall. Below it are the only two dials a camera actually has: focal length and distance to the subject. The instrument renders exactly what the sensor would catch, from the pinhole projection apparent size = f · (real size / distance) — nothing hidden, nothing faked. Start by moving only the focal length. The frame tightens; the whole picture zooms. But the one number that says how much the far person is dwarfed by the near one — the near : far size ratio — does not so much as flinch.
Near : far size ratio
3.50 ×
= (d + Δ) / d — no focal length in it
Angle of view (horiz.)
39.6°
50 mm on full-frame
Background
looms
far subject fills 13% of frame height
16 mm (super-wide) to 360 mm (long tele). This only sets the angle of view — how much of the world fits in the frame.
Step in or back up. This is the dial that owns perspective — watch the near : far ratio move the instant it does.
With lock on, the two dials are chained by Irmin Roberts's rule so the near subject stays exactly the same size — the lens is doing precisely its job — and yet the background still swells or collapses. That isolates distance as the sole cause. It is the shot Hitchcock built for Vertigo (1958).
That single fraction is the whole argument. The near person and the far person are the same height; the near one appears larger purely because it is closer, by exactly the ratio of the two distances, d_far / d_near = (d + Δ)/d. Focal length multiplies both figures by the same amount, so it slides straight out of the ratio. Change the lens and you change how much of the scene is cropped in; you do not change the perspective one part in a thousand. Change the distance and the perspective is a different picture.
Now tick lock subject size. The dials chain together — step back and the focal length lengthens to keep the near subject filling the frame, step in and it shortens. This is the sophisticated objection made operable: "fine, but on a real shoot the tele obviously compresses — I can see it." Here the lens is explicitly compensated, the subject is pinned to one size, and the only thing left free is distance. Press Run the Vertigo dolly and watch the background alone rush toward you or fall away. The subject never moves; the world behind it does. That is the dolly zoom, and it is a proof, not a trick: with subject size held constant, distance is the only variable left to blame.
The check — every number recomputed in front of you
Hold the distance fixed at d = 8 m (so the far subject is at 28 m) and read off three lenses from super-wide to long tele. Angle of view collapses from 74° to 10°; the near and far figures balloon; and the near : far ratio — the green column — does not move. The green column is the cross-check, not a claim.
| focal | angle of view | near fills | far fills | near : far |
|---|---|---|---|---|
| 24 mm | 73.7° | 21.9% | 6.25% | 3.50 × |
| 85 mm | 23.9° | 77.5% | 22.1% | 3.50 × |
| 200 mm | 10.3° | 182% | 52.1% | 3.50 × |
For the current dials, the ratio and — when locked — the dolly-zoom focal length, plugged through live:
The dolly-zoom constraint is Roberts's own: d = w / (2 · tan(½ · FOV)), which for our framed height w = 3.0 m on a 24 mm sensor gives f = 24·d / w = 8·d — the subject stays 58% of the frame at every distance while the background magnifies as d / (d + Δ). Run it yourself: node research/the-compression-that-isnt-the-lens/verify.mjs.
So what is true and what is folklore? The folk claim comes in two halves — "telephotos compress distance" and "wide-angles distort faces" — and both halves are the same distance effect wearing a lens's name. You reach for a long lens to shoot a portrait, so you back up; backing up flattens the near : far ratio, and the world stacks up behind the face. You reach for a wide lens in a small room, so you step in; stepping in exaggerates the ratio, and the nose looms. The lens did not do either. It set your framing, the framing forced your distance, and the distance did the rest.
What's exactly true, what's idealised, and where the popular story runs ahead
Exactly true
Perspective — the projected relative sizes and overlaps of things — is a function of the viewpoint alone. For two objects at distances d₁ and d₂, the apparent-size ratio is d₂ / d₁, with no focal length in it; this is the exact pinhole projection size ∝ f·(H/d) with the common f cancelled. Wikipedia's article on perspective distortion states it plainly: "linear perspective changes are caused by distance, not by the lens per se." The dolly-zoom attribution is firm too: the effect was devised by Irmin Roberts, a Paramount second-unit cameraman, for Alfred Hitchcock's Vertigo (1958), holding subject size fixed while the background scale changes.
The crop-equivalence, and its fine print
Because focal length only selects an angular window from a fixed viewpoint, a wide-angle frame cropped to a telephoto's field of view gives geometrically the same perspective — the same near : far relationships. That equivalence is geometric only. The crop throws away resolution (fewer pixels on the subject), and "the same photo" quietly ignores depth of field and bokeh, which do shift with focal length and aperture and are part of the look people call "compressed." The instrument draws the ideal rectilinear projection; it does not model those.
Perspective distortion is not optical distortion
This is the landmine. Lenses genuinely do distort — barrel distortion in wide and fisheye glass, pincushion in some teles, moustache distortion in zooms. Those are real optical aberrations, separate from perspective. An extreme wide-angle close-up bulges a face from both the distance ("extension") effect and barrel distortion at once. So the precise claim is not "lenses cause no distortion" — they do. It is narrower and exact: the lens does not cause perspective compression. This page isolates the perspective term; the optical term is a different, real thing.
Where the popular story runs ahead of the record
The honest boundary: focal length does not cause perspective compression — but it is not irrelevant. It sets the angle of view, and the angle of view dictates the distance a photographer chooses (you back up for a tele portrait), and that distance is the cause. Focal length is correlated with compression, not causally upstream of it. The popular story runs ahead of the evidence only in assigning cause to the glass; the correlation it notices — long lens, compressed look — is completely real, because photographers really do stand far back when they reach for a long lens.
Idealised / representative / free choices
- Free choice: the two subjects are Δ = 20 m apart; the sensor is full-frame 36 × 24 mm; the locked framing keeps a 3.0 m tall slice in view (subject at 58% of frame). Change the numbers and the values move; the focal-length-independence of the ratio does not.
- Idealised: an ideal rectilinear (pinhole) lens with no optical distortion, a point aperture (infinite depth of field), and figures treated as flat cut-outs facing the camera. Real glass adds the aberrations named above.
- Representative: the subjects are 1.75 m tall, the camera 1.5 m high. These set where things sit in the frame, not whether the ratio depends on focal length — it does not, for any choice.
The deeper principle is shared with a very different craft. In The Floor That Builds Itself, Renaissance linear perspective is shown to be forced geometry set by the viewing distance — the same truth from the painter's side of the picture plane. Fix the viewpoint and the projection is fixed. Everything else is a choice about how much to crop.