How animation tricks your eye

Hold a hand in front of your face and wave it slowly. Now wave it fast. At some point it stops being a hand in a series of positions and becomes a blur with a hand somewhere inside it.

Your visual system has a frame rate problem, and it solves it by lying to you in a very particular way. Every moving image you've ever watched is built on that lie — twenty-four still pictures a second, each one motionless, adding up to something your brain insists it can see moving.

The explanation you almost certainly got for how that works is wrong. It's been wrong for about a century, it's still printed in textbooks, and the correct answer is more interesting.

Persistence of vision is not the answer

The standard story goes like this: an image lingers on your retina for a fraction of a second after it's gone, so consecutive frames overlap and smear into continuous motion. Neat, mechanical, easy to remember, and it doesn't survive five minutes of thinking about it.

Persistence of vision is real. Look at a bright light and shut your eyes and you'll see the afterimage. Wave a sparkler and you'll draw a line in the air. That's a genuine property of the retina and it takes something like a twentieth of a second to fade.

But look at what it would actually produce. If each frame hung around while the next arrived, you wouldn't see motion — you'd see a pile-up. A smeared superimposition of a runner in six positions at once, getting muddier with every frame. That isn't what anybody sees at the cinema.

Persistence of vision explains why you don't see the black gaps between frames. It doesn't explain motion at all. The two problems got welded together in the nineteenth century and stayed welded, largely because Peter Mark Roget — the thesaurus man — presented a paper in 1824 about persistence of vision and moving objects, and the phrase was catchy enough to outlive the reasoning. Joseph and Barbara Anderson published a thorough demolition of it in 1993, under the entirely reasonable title The Myth of Persistence of Vision Revisited, and the myth carried on anyway.

What's actually happening has a different name

The effect is called apparent motion, and it was pinned down by Max Wertheimer in 1912 in a paper that ended up founding Gestalt psychology.

Wertheimer's setup was crude and the results weren't. Flash a light in one place, then flash a second light a short distance away. Vary the gap between the flashes and the experience changes in stages.

With a long gap, you see two separate lights, one after the other. Nothing moves. With a very short gap, you see two lights on at once. Also nothing moves. Somewhere in between — around fifty to a hundred milliseconds, depending on the distance — you see one light travelling from the first position to the second. It's not a smear or an inference. It looks like movement, as vividly as real movement looks like movement, and you can't switch it off by knowing better.

Wertheimer separated two versions. Beta movement is what most animation runs on: you see the object itself move across the gap. Phi is stranger, appearing at faster rates, where you perceive pure movement in the space between without the object appearing to travel through it — a sense that something happened there without anything visible doing it. In casual writing the two names get swapped constantly, including by people who should know better.

The important part is what this reveals. Your brain isn't recording motion; it's constructing it. Motion is a computed property, generated by dedicated neural machinery that takes changes in position over time as input, and that machinery can be fooled by inputs that never actually moved. Animation isn't sneaking past your perception. It's feeding your perception exactly what it's built to eat.

The Victorians got there without knowing why

The devices arrived decades before the explanation, which is the usual order of these things.

The thaumatrope, popularised around 1825, was a disc with a bird on one side and a cage on the other, spun on strings until the bird appeared to be inside. Joseph Plateau's phenakistiscope of 1832 used a slotted disc of drawings spun in front of a mirror — you looked through the moving slots at the reflected drawings, and the slots did the essential work of showing you one drawing at a time rather than a blur.

Then the zoetrope, patented in its practical form in the 1830s by William George Horner and sold widely from the 1860s. A drum with slots cut in the sides and a paper strip of drawings around the inside. Spin it, look through the slots, and a horse gallops. Several people can watch at once, which the earlier toys didn't allow.

Charles-Émile Reynaud's praxinoscope of 1877 replaced the slots with a ring of mirrors at the centre, which is cleverer than it sounds — the slot method throws away most of the light, and mirrors don't.

Every one of these gadgets solves the same engineering problem: show a sequence of static drawings, and make sure each one is visible on its own rather than sliding past as a continuous streak. Get that right and the motion appears for free, because your brain supplies it.

Flicker is a separate problem with a separate fix

Twenty-four frames a second is enough for smooth motion. It is nowhere near enough to stop the picture appearing to pulse.

The threshold at which a flashing light stops looking like flashing and starts looking steady is the critical flicker fusion frequency, and for most people in normal conditions it's somewhere around fifty to sixty flashes a second. Brighter light pushes it higher. So does peripheral vision, which is why a screen you're not looking at directly can flicker visibly while the same screen looks fine head-on. Twenty-four is far below that line, and early audiences complained bitterly — "the flickers" became a nickname for the cinema.

The fix is beautifully cheap. Film stock costs money; light doesn't. So projectors were built with shutters carrying two or three blades, which interrupt the beam more than once per frame. Each frame gets flashed twice, or three times, in exactly the same position on the screen. Twenty-four frames become forty-eight or seventy-two flashes, comfortably above fusion, and not a single extra frame was shot.

Which is a clean demonstration that the two effects are independent. Motion perception is handled by frame rate. Flicker is handled by flash rate. You can raise one without touching the other.

Why twenty-four, of all numbers

There's nothing perceptually special about twenty-four. It's an accounting decision that hardened into an aesthetic.

Silent films weren't shot at a fixed rate at all. Cameras were hand-cranked, typically somewhere between sixteen and twenty frames a second, and the rate drifted with the operator's arm. Projectionists ran films faster or slower to fit the schedule. This is why silent footage transferred at modern speeds has that hurried, scuttling quality — it's being played too fast, and the people in it were not actually like that.

Sound ended the improvisation. An optical soundtrack printed along the edge of the film has to pass the reader at a constant, sufficient speed to carry usable audio frequencies. Too slow and the sound is muddy. Studios wanted the slowest rate that gave acceptable sound, because every extra frame per second meant more stock through the camera, more stock through the printer, and more stock in every print shipped to every cinema. Twenty-four was the compromise that got settled on in the late 1920s and it's been the standard ever since.

So the "cinematic" frame rate is a bill from a film laboratory in 1927.

Motion blur is doing more work than the frames are

Here's the piece that gets left out, and it explains most of what people mean when they say something "looks like film".

A film camera doesn't take instantaneous snapshots. Its rotating shutter is open for a portion of each frame's duration — conventionally half, the so-called 180-degree shutter — so at twenty-four frames a second each frame is exposed for about a forty-eighth of a second. Anything moving quickly is smeared within that single frame. A whirling wheel spoke becomes an arc. A punched fist becomes a comet.

That smear is not a defect. It's a huge quantity of extra motion information stuffed into a still image, and it gives the visual system the direction and speed cues it needs to stitch the sequence together comfortably.

Take it away and things break. Shoot with a very fast shutter — the technique used for the beach sequences in Saving Private Ryan in 1998 — and each frame is razor-sharp with no smear at all, and the result is a horrible jittery staccato that feels like the world is skipping. That harshness was chosen deliberately, and it works precisely because it withholds what your eye expects.

Computer-generated animation has the opposite problem. A virtual camera has no shutter, so every render is perfectly sharp by default and the motion strobes badly. Motion blur has to be calculated and added on purpose, and getting it right is a substantial part of why CG sequences look like they belong in the same film as the live action.

Ones, twos and threes

Hand-drawn animation almost never uses twenty-four drawings a second, because twenty-four drawings a second is ruinous.

Working "on twos" means drawing twelve pictures a second and photographing each one for two frames. It looks fine. Motion perception doesn't need more, and the doubled frames don't introduce flicker because the projector's shutter is dealing with that separately. Almost all classic Western animation is on twos as its default state.

The exceptions are where the craft lives. Fast action goes on ones, because at twelve a second a quick movement travels too far between drawings and the eye stops linking the positions — the object appears in two places rather than moving between them. Slow drifts can go on threes or fours without anyone noticing. Skilled animators change the value inside a single shot, and audiences never spot the switch.

Japanese television animation leaned much harder on this for economic reasons, frequently working on threes — eight drawings a second — and compensating with camera moves, held poses, and elaborately detailed single frames. That constraint produced a whole visual language: the long held reaction shot, the sliding pan across a still background, the pause before impact. What began as a budget limit became a style, and it's now used deliberately by studios that could easily afford more drawings.

American television animation from the late 1950s went further still, breaking characters into reusable layered parts so only a mouth or an arm needed redrawing while the rest of the cel sat still. It's very obvious once you've seen it. It also made a daily animated series financially possible for the first time.

Keys, breakdowns and the people in between

Nobody draws an animated sequence in order from the first frame to the last. That approach exists, it's called straight-ahead animation, and it's used for fire, smoke, water and anything else where spontaneity matters more than control. For characters it's a disaster, because the drawing drifts and by frame sixty the face belongs to someone else.

The standard method is pose to pose. A senior animator draws the extremes — the poses at the ends of each movement, where the action changes direction or intent. Then the breakdowns, the crucial in-between positions that determine the path the movement takes rather than just its endpoints. Only then does anyone draw the inbetweens that fill the remaining gaps, and historically that was a separate and more junior job, with the assistant working over a lightbox from the animator's drawings and a numbered chart specifying exactly how the gaps should be spaced.

The spacing chart is where the timing actually happens, and this is the distinction that trips people up. Timing is how many drawings a movement takes. Spacing is how far apart they sit. A movement with drawings bunched up at both ends and stretched out in the middle appears to accelerate away and decelerate into its destination, which is what real objects with mass do. Evenly spaced drawings produce the dead, floating, mechanical motion that instantly marks out amateur work.

Ollie Johnston and Frank Thomas set out the whole toolkit in The Illusion of Life in 1981 as twelve principles, and they're still taught essentially unchanged. Squash and stretch, which deforms a shape under acceleration and impact while conserving its apparent volume. Anticipation, the small backward movement before a forward one. Follow-through and overlapping action, where trailing parts arrive late. Slow in and slow out, which is the spacing point above. Arcs, because almost nothing biological travels in a straight line. They're not stylistic preferences. They're a compressed description of how mass and inertia behave, discovered empirically by people who spent their lives watching movement frame by frame.

Tracing over the real thing

Max Fleischer patented rotoscoping in 1917 — filming a live performer, projecting the footage one frame at a time onto a drawing surface, and tracing over it.

The results are peculiar, and the peculiarity is instructive. Rotoscoped movement is accurate and often looks wrong. It carries all the small corrections, wobbles and hesitations of a real body, and animation drawn by hand deliberately removes those in favour of exaggerated arcs and clean holds. Traced motion frequently reads as floaty or uncanny next to drawn motion, which is why studios have generally used rotoscoping as reference material to study rather than as a line to trace directly. The Disney artists shot live-action footage extensively while making Snow White and the Seven Dwarfs in the 1930s and used it mainly to understand weight and balance.

The technique never went away. Ralph Bakshi used it heavily in the 1970s, and Richard Linklater's A Scanner Darkly in 2006 used a digital interpolated version to produce a permanently shifting, unstable surface over recognisable performances. Motion capture is the same idea with the drawing step removed and the same fundamental difficulty intact: recorded human movement mapped onto a non-human proportion tends to look subtly ill.

Why higher frame rates feel wrong

More frames should be better. Smoother motion, less strobing on pans, sharper detail in movement. Audiences hate it.

The Hobbit: An Unexpected Journey was shown at forty-eight frames a second in 2012 and the reaction was blunt and widespread — viewers described it as looking like television, or like a behind-the-scenes video, or like watching actors in costumes standing on a set. Which is roughly what it was, rendered with enough temporal resolution to show it.

Television sets do this to films on their own with motion interpolation, generating synthetic frames between the real ones. It's usually on by default and it's known as the soap opera effect, for the obvious reason.

The honest explanation is that there's no perceptual law protecting twenty-four. Higher rates aren't objectively worse. They're unfamiliar, and a century of association has trained everyone alive to read a particular combination of frame rate and motion blur as "story" and a smoother one as "record of real events happening in front of a camera". Strip away the blur and the slight judder and the illusion of a world gives way to the visible fact of a production.

Whether that association is permanent is an open question. Nobody in 1927 thought twenty-four frames looked filmic either. It just looked like what films looked like, which in the end is the same thing arriving from the other direction.