Dolly and zoom

Film has had precise words for this for a century, and interface work keeps reinventing vaguer ones. A dolly moves the camera itself through the scene, which changes where the viewer is standing. A zoom changes the lens, which scales the picture without the viewer going anywhere. In a single frame the two are almost indistinguishable, since both make the subject bigger. In motion they are nothing alike, because a dolly changes the relationship between near things and far things and a zoom cannot: walk toward a doorway and the wall beside you sweeps past while the hills behind barely move, but zoom in on the same view and the doorway, the wall and the hills all grow by exactly the same factor. The rest of the family is worth knowing for the same reason: truck slides the camera sideways, pan and tilt rotate it in place, and pedestal raises it.

The famous combination is the dolly zoom, invented for Hitchcock’s Vertigo and named after it ever since. Dolly in and zoom out at matched rates and the subject holds its size exactly while everything behind it falls away, or run it the other way and the background looms in. Nothing in the frame is moving except the space itself, which is why the shot reads as dread rather than as motion. In an interface the same trick has a plainer job: it is the strongest available way of saying you moved rather than the content changed, because holding one element still while its surroundings deform is something only a change of viewpoint can do.

This is exactly where it must not be confused with parallax, which is the other way to imply depth and is a different claim. Parallax moves layers at different rates as the reader scrolls, so depth is inferred from how fast each layer travels sideways. A dolly zoom changes the field of view while the camera moves, so the subject is held and the space around it warps, and the depth is inferred from how differently near and far things scale. One is about speed, the other about scale.

Practically, a zoom is cheap and a dolly is not. A zoom is one transform on one container. A dolly needs the scene to have real depth: either per-layer scale factors derived from each layer’s distance (near layers change most, far layers barely at all), or a genuine perspective on the container with each layer at its own translateZ, which lets the browser do the arithmetic. The 3D route is usually the better one, since it keeps the depths as one number per layer rather than as a table of scales someone has to keep consistent. Two cautions. The vocabulary only pays off if the space is stable, which means a card that dollied in has to dolly back out along the same path, and a scene that rearranges between moves has forfeited the metaphor. And strong camera moves are a vestibular trigger, more so than anything else in this category, so a push-in belongs behind prefers-reduced-motion with a cut or a fade as its alternative. The mildest member of the family, a slow drift across a still image, has its own name: the Ken Burns effect.