← Blog

Right Brain · Reading

Why You Recognise People by the Way They Walk

You have had this experience. Someone is far enough away that their face is a smudge, the light is bad, they are wearing a coat you have never seen — and you know exactly who it is before you could say why. Not from a feature. From the way the whole thing moves. This turns out to be one of the most thoroughly studied phenomena in perception, and one of the strangest. Take in the whole. Before the parts.

By ComunAI · 17 August 2026 · ~6 min read

Twelve dots

In 1973, Gunnar Johansson attached small lights to the major joints of a person, filmed them in a dark room, and showed the resulting displays to observers. Nothing else was visible: no body, no outline, no face. Just a scatter of bright points against black.¹ Strong

Frozen, a single frame of one of these displays is meaningless — people shown a still frame typically describe an arbitrary constellation of dots, and cannot say what it is. Set the same dots in motion and a walking human appears immediately and irresistibly, usually within a fraction of a second. Not a figure that observers work out. A figure they cannot help seeing.

Johansson's displays became known as point-light walkers, and they have been the workhorse of this field for half a century, precisely because they strip away everything that could be doing the work except the motion itself.

How much survives the stripping-down

The surprising part is not that observers see "a person". It is how much detail comes through a dozen dots.

• Who it is. James Cutting and Lynn Kozlowski showed point-light displays of people to observers who knew them, and found they could identify individual friends from gait alone, with the face, body shape and clothing removed.² Moderate Small samples and modest accuracy — well above chance, nowhere near perfect.
• What kind of person. In a companion study, observers could judge the sex of a point-light walker at better than chance, and the researchers traced the available cue partly to the relative movement of shoulders and hips.³ Moderate
• What they are doing. Walking, running, dancing, climbing, carrying something heavy — actions are read off these displays readily, which is why the technique spread so far beyond its original question.

Nikolaus Troje later showed that gait patterns can be decomposed into components and resynthesised, so that a walker can be manipulated systematically to test what observers are actually using — turning a demonstration into a measurement tool.⁴ Moderate

It is not just "you are good at motion"

An obvious deflation would be that this is nothing special — the visual system is good at motion, and a walker is motion. The evidence pushes back on that in two ways.

It is remarkably robust to noise. Peter Neri, Concetta Morrone and David Burr compared the detection of biological motion against simpler motion tasks and found that observers could pull a walker out of a field of distractor dots with a sensitivity that ordinary motion integration does not straightforwardly predict.⁵ Moderate

It breaks when you turn it upside down. Invert a point-light walker and recognition degrades badly — far more than inverting most patterns does.⁶ Moderate That signature should look familiar: it is the same disproportionate inversion cost that marks face perception, which we went through in Why You See Faces in Things. In both cases the reading is the same — the system is not assembling the percept from independent parts, because a process built from independent parts would not care much which way up they were.

Where in the brain, and how early in life

Emily Grossman and colleagues used fMRI to look for regions responding more to point-light biological motion than to scrambled control displays, and found a focus in the posterior superior temporal sulcus.⁷ Moderate Imaging shows correlation, so the stronger result is the follow-up: applying transcranial magnetic stimulation over that region impaired observers' ability to detect biological motion, while a control site did not.⁸ Moderate That moves the claim from "this area is active" toward "this area is doing something the task needs".

And the developmental end is stranger still. Francesca Simion, Lucia Regolin and Hermann Bulf tested newborns — two days old, essentially no visual experience — and found they looked longer at a biological-motion display than at a scrambled one.⁹ Early One infant study with the usual limits of the looking-time method, and it should be graded modestly. But it is at least consistent with the sensitivity being built in rather than entirely learned.

Randolph Blake and Maggie Shiffrar's review is the place this literature is gathered and weighed, if you want the whole picture rather than the highlights.¹⁰

Where the evidence stands

• A dozen moving dots produce an immediate, compelling human figure. Strong — replicated continuously since 1973.¹
• A single static frame carries almost none of it. Strong — the defining feature of the phenomenon.
• Observers can identify friends, and judge sex, from gait alone. Moderate — above chance, small samples, modest accuracy.^2,3
• Inverting the display disrupts it disproportionately. Moderate⁶
• Posterior STS is causally involved. Moderate — imaging plus a TMS result.^7,8
• Newborns already prefer biological motion. Early — one looking-time study.⁹
• You can be trained to read gait reliably enough to identify strangers. No evidence — nothing in this literature supports that, and we are not going to imply it.

What this means for Right Brain

Right Brain is named for a metaphor, not a hemisphere — the open, whole-first way of looking, as against the narrow, part-by-part one. Biological motion may be the single cleanest case anyone has found for why that distinction is not decorative.

Consider what a part-by-part system would have to do here. Locate each dot. Track it. Compare its trajectory to the others. Infer a skeleton. Match the skeleton to a body. It would be slow, and it would be brittle in noise, and it would work about as well upside down as right way up. What actually happens is none of that: the figure arrives whole, in under a second, and falls apart the moment you invert it. The pattern is the unit. The dots are not.

That is the same lesson depth taught us in Why Flat Pictures Look Three-Dimensional — crop a scene to a patch and the depth vanishes, because it was never in the patch — running here in the dimension of time rather than space. And it is worth pairing with Attention vs. Perception, because a walker in a crowd is a case where the two questions come apart: your visual system can extract the motion long before you have decided to look at anyone.

The perception games in the app are that idea shrunk to a few quiet minutes: no score, no verdict about your abilities, just your own inference machinery caught briefly in the act. The standing limit applies as always — there is no good evidence that playing perception games improves your everyday seeing. Broad transfer is the weak link in this whole field, and we covered it properly in Do Brain-Training Games Actually Work? Contested