the legs learn a moving world so well that the learning is only visible on the day the motion stops
R780 closed with a watcher who knows a planet only because light that was always arriving briefly wasn't. There is a bodily version of the same trick, and its instrument is not a telescope but a pair of legs. Step onto an escalator that has stopped — one you can plainly see is stopped — and for a step or two the body lurches, as if the ground had cheated. Nothing was taken away in that moment except motion. The oddness is the measurement: it is how you find out your legs had been expecting the motion all along.
Reynolds and Bronstein measured it in 2003 with a sled instead of an escalator: a platform driven by a linear motor at 1.2 m/s, walked onto by fourteen volunteers whose speed, trunk movement, and leg-muscle activity were recorded. First they walked onto the still sled at their ordinary boarding speed, about 0.60 m/s. Then came twenty rides onto the moving sled, and boarding speed climbed to about 0.90 m/s — the body had quietly rewritten the step to meet a moving world. Then the sled was stopped, the volunteers were told plainly and unequivocally that it would not move, and they walked on again. They still boarded at about 0.71 m/s: too fast for the still platform, with a large forward overshoot of the trunk and extra leg-muscle activity. They were, by the report, surprised by their own behaviour. By the second still trial, most measures had settled back near baseline.
Run the sled trials. In order: board the still sled, take the twenty moving rides (compressed here into one press), then — having been warned it is stopped — board it still again, twice. Watch the boarding speed, not the sled.
Phase: before any training. The sled is still.
Boarding speed
No trial yet. Ordinary boarding in the study: ~0.60 m/s.
Note where the surprise lives: not in what you were told — you were told, and you believed it — but in a speed chosen before your foot even touched the platform.
The narrowings, so this page does not overclaim. First, the numbers are one study's means for one sled speed: 1.2 m/s is a fast walkway, and a slower platform, a different warning, or a different history would give different figures. Second, the after-effect is brief — most of it gone by the second still trial — so this is a fast, local recalibration, not a lasting false belief. Third, nobody in the study was fooled about the facts: the volunteers saw the still sled and accepted the warning. What failed to update was not their knowledge but a motor habit running underneath it.
That last narrowing is the page. The transit watcher of R780 knows by interruption and knows that it knows: the dip is written in a notebook. Here the interruption reveals a knowledge nobody can write down, held in a system that does not take instructions from the part that does the knowing. The legs had a complete, working theory of the escalator — approach fast, brace forward, expect to be carried — and the theory stayed invisible for as long as the world agreed with it. It became visible only in the one step where the motion that was always there briefly wasn't, and the body, faithful to its twenty rides, stepped for the world it had learned instead of the world in front of it.