Will it take off?

A plane stands on a conveyor belt as long as a runway, and the belt runs backwards exactly as fast as the plane rolls forwards. People have been arguing about what happens next for more than twenty years.

Scroll down to find out
Airspeed
–
Wheels
–
Belt
–

Drag to look around.Drag sideways to look around.

What pushes a plane

It takes off, almost exactly as it would from an ordinary runway. If that sounds wrong, the instinct behind it is a sound one, and it’s worth seeing where it goes astray.

Run on a treadmill and you stay exactly where you are. Put a car on a rolling road, the kind a garage tests it on, and you can drive it at 200 km/h without it moving an inch. In both cases the belt cancels the motion. A plane held in place like that would never fly, because a wing lifts only when air flows over it, and on a still day enough air flows over it only if the plane moves. That part of the argument is right.

The mistake comes a step earlier. A runner and a car move by pushing on the ground beneath them: feet and tyres shove the belt backwards, the belt shoves them forwards, and a belt that runs away underneath takes that push with it. A plane doesn’t push on the ground at all. Its propeller grips the air and throws it backwards, and the air, pushed, pushes back.

Propeller’s push
–
Slipstream
–
The Cessna’s propeller in the live wind tunnel as the plane starts its run, drawn as the disc it sweeps, with the plane outlined behind it for scale. Smoke shows the air and colour its pressure, blue below normal and amber above. The air ahead is drawn in and thrown back; the push it takes is the push the plane gets. The readouts are the real Cessna’s.

Nothing in that exchange touches the ground. The engine doesn’t drive the wheels. They turn freely, and they’re there only so the plane doesn’t scrape along on its belly. The belt can spin them as fast as it likes: it is pushing on the one part of the plane that isn’t doing any pushing.

Still, the wheels are where plane and belt meet, and that’s where most people who say no dig in. So it’s worth looking closely at a wheel.

The wheels just turn

The belt can reach the plane only through its tyres, so everything depends on what a freely turning wheel passes on.

Plane’s speed
1×
Wheel’s spin
1×

Here is one of the Cessna’s main wheels rolling along an ordinary runway, slowed right down. Three points on the tyre leave trails over the ground, and each traces a row of arches that come down to a sharp point. At that instant the bottom of the tyre is standing still on the runway, and the wheel is pivoting about it.

The arrows show how fast each part of the wheel moves over the ground, in multiples of the plane’s speed. The axle keeps pace with the plane, the top of the tyre goes twice as fast, and the bottom, for that one instant, not at all.

Now roll the wheel onto the belt, running backwards as fast as the axle moves forwards. The bottom of the tyre has to travel backwards with the belt, so the wheel spins twice as fast and the arches open out into loops.

The axle, which carries the plane, goes on along the same straight line at the same speed. The belt has changed how the wheel turns, not where it goes.

That leaves one question: how hard can the belt push the plane back through a free wheel? Here are the forces at the start of the run, drawn to scale.

The propeller pulls with 2.6 kilonewtons. The tyres flex as they roll, which costs 0.23 kilonewtons, on the belt just as on a paved runway. The belt adds only the push it takes to spin the wheels up twice as fast: 0.03 kilonewtons, about one percent of the propeller’s pull.

Run the belt three times as fast and the wheel spins four times as fast, in tighter loops. The axle’s line still doesn’t budge. Whatever the belt does, all it changes is how fast the wheels turn.

Lock the wheels and everything changes. A plane with its brakes on is a crate sitting on the belt, and the belt carries it backwards. That is the one honest way to keep this plane on the ground, though it would keep it on an ordinary runway too.

The propeller, meanwhile, doesn’t care what the wheels are doing. It holds on to the air, which isn’t going anywhere, and pulls.

Lift reads only the air

So the plane rolls down the belt as it would down a runway. Here is the whole takeoff with real numbers, for the Cessna 172: the four-seater many pilots learn on, and the most-built aircraft in history.

At its maximum weight of 1,157 kilograms, from a paved runway at sea level, it leaves the ground after 293 metres, about 18 seconds after the brakes come off, at 106 km/h through the air.Cessna, Model 172S Pilot’s Operating Handbook, §5, short-field takeoff at 2,550 lb, sea level, 15 °C: a ground roll of 960 ft, lifting off at 51 knots on the airspeed indicator, which the handbook’s calibration table puts at about 57 knots through the air. Its propeller starts by pushing with about 2.6 kilonewtons, a little under a quarter of the plane’s weight, and pushes less as the plane speeds up.Our estimate: the thrust that, after drag and the tyres’ rolling resistance, reproduces the handbook’s ground roll. Every number about the takeoff on this page comes from that one model, which also flies the plane in every figure.

Put it on the belt and the numbers that matter don’t move. The plane still needs 106 km/h through the air, and the propeller still pushes on the same still air. Only the wheels notice: they finish the run turning at 211 km/h instead of 106, and spinning them twice as fast costs about one percent of the propeller’s push. The plane lifts off 3.5 metres and a fifth of a second later. The belt does have to be as long as the runway.

Run the belt faster still and the plane flies anyway, a little later each time. The tyres are the first thing to suffer. The best ones Cessna lists are rated for 257 km/h, and the wheels reach that at lift-off once the belt runs at about 1.4 times the plane’s speed.Main tyres 6.00-6, nose tyre 5.00-5 (handbook, §8). The premium tyres in its equipment list (§6) are rated for 160 mph; Goodyear’s Flight Custom III 6.00-6, a common choice, weighs 5.1 kg.

None of this reaches the wing. To see what does, put the wing in a wind tunnel, the same live one as on our page about lift.

Air past the wing
–
Floor past the wing
–

This is a live wind tunnel, the kind engineers test wings in: the wing stays put and the air streams past it, here at the takeoff speed of 106 km/h. It’s a slice through the Cessna’s wing, two metres above the floor, and the floor is the runway. Seen from the wing, the runway slides back at exactly the speed of the air, because in still air the ground and the air are both standing still.

Smoke shows the air, colour its pressure, blue below normal and amber above, and the arrow the lift, scaled so that here, on the runway at takeoff speed, it reads as the plane’s whole weight.

Now make the floor the belt, sliding back twice as fast. Up at the wing, the smoke, the pressure and the lift don’t change at all.

A moving floor can move air only by dragging it along. Over the length of a takeoff run, a real belt drags a layer of air a metre or two deep, but slowly at the top: about half the belt’s speed a couple of centimetres up, a tenth of it a metre up. And it drags that air backwards, so wherever it reaches the wing at all, it adds a whisker of headwind. This tunnel’s floor leaves the layer out and only stops the air passing through it.

Stop the floor altogether and keep the air blowing: a plane tied down in a 106 km/h storm. The lift is the same again. Wind alone can lift a plane that hasn’t moved, which is why parked planes are tied down.

And the other way round: the floor racing, the air still. That’s a plane held in place, the version of the puzzle we come to next. No air goes past the wing, so there’s no lift, whatever the floor does.

The wing reads only the air going past it. That is why pilots take off into the wind whenever they can: for a Cessna, every 9 knots of headwind takes a tenth off the takeoff run.Handbook, §5, note 3 to the takeoff tables. It is why aircraft carriers steam into the wind before they launch. And it is why many of the wind tunnels that car makers and racing teams use have a moving belt for a floor, running at the speed of the air: a fixed floor would drag the air in a way no road ever does.The belt’s dragged layer above is a rough estimate: a turbulent boundary layer growing over the few hundred metres of belt ahead of the plane. On moving floors in car wind tunnels see, for example, S. Wang, T. Avadiar, M. C. Thompson and D. Burton, “Effect of moving ground on the aerodynamics of a generic automotive model: The DrivAer-Estate”, Journal of Wind Engineering and Industrial Aerodynamics 195, 104000 (2019).

The version where it can’t

There is one way to read the puzzle in which the plane really does stay put, and it’s the version most of the people who say no have in mind.

Suppose the belt is meant to match not the plane’s speed but the wheels’ own, as a speedometer on the wheel would read it. The trouble, as the wheel showed, is that a wheel rolling on a belt turns at the plane’s speed plus the belt’s. Speed the belt up to match the wheels and the wheels speed up by just as much, so the belt can never catch them unless the plane isn’t moving at all. In this version the plane stands still by definition. The only question left is whether any belt could make it.

In principle one could, and only through the wheels’ inertia. A wheel resists being spun up, so the belt has to push it round, and if the belt speeds up hard enough, those pushes on the three tyres add up to the propeller’s pull. For a Cessna’s small wheels that takes a belt accelerating at 17 g, about twice what a fighter pilot can stand.Our estimate, from the wheels’ inertia: each main wheel with its tyre resists spinning up like 5.4 kg at its rim, the nose wheel like 3 kg. Friction in the bearings would help the belt a little, by a few percent.

Time
–
Belt speed
–
Belt accelerating at
–
The belt doing whatever it takes to keep a Cessna at full power standing still. Its only grip is through the tyres, so it spins the wheel ever faster, and its push back exactly matches the propeller’s pull. Slowed ten times.

Within 0.42 seconds the belt passes the tyres’ rated speed, and in about 2 seconds the speed of sound. Long before then the tyres would come apart, the plane would drop onto its wheel rims, and the belt would drag it away like any other crate. So in this version the plane really doesn’t take off. But the belt didn’t beat the propeller. It beat the wheels, by destroying them.

Why the argument never lands

The puzzle is older than most of the people arguing about it would guess, and no answer has ever ended it.

The earliest copies anyone has found are on Russian forums in the summer of 2003: one on avia.ru, an aviation forum, dated 24 July, and another on iXBT, a computer forum, posted on 4 August by someone whose fellow students were already fighting over it. Nobody knows who first asked it.The avia.ru post, signed StealthMan, is cited in Wikireality’s account of the puzzle. The iXBT thread was traced by the man who first posted the puzzle in English, as told in A. Baio, “Origins of the ‘Plane on a Treadmill’ Meme”, waxy.org (2008), and its updates.

In July 2005 a half-Russian reader carried it to PhysOrg, an English-language physics forum. By November it had reached pilots, splitting a thread on Flightinfo, a forum for professional pilots, and an AVweb columnist was writing about the quarrel spreading through the pilot community. In February 2006 Cecil Adams took it up in The Straight Dope, the newspaper column in which he answered readers’ questions, and said the plane takes off. A month later he was back with a column that opened with his readers’ verdicts: “My confidence in Cecil has taken a gigantic hit.” One of them, an engineering student, swore that a belt accelerating hard enough would hold the plane where it stood, which, as we’ve seen, is true only of the belt that destroys the tyres. By December the technology columnist David Pogue was writing in The New York Times that the puzzle was “ripping around the Internet”, and that on a single trip to Canada he had overheard it in three airport conversations.Baio (2008). C. Adams, “An airplane taxies in one direction on a moving conveyor belt going the opposite direction. Can the plane take off?”, The Straight Dope, 3 February 2006, and his follow-up in the Chicago Reader, 2 March 2006. D. Pogue, “The Airplane-Treadmill Conundrum”, Pogue’s Posts, The New York Times, 11 December 2006.

The MythBusters set out to settle it with an experiment, broadcast in January 2008. Jamie Hyneman thought the story wasn’t worth doing, since of course the plane takes off; Adam Savage won him round by printing out about two inches of the internet’s arguments. The rig, which Savage still counts among his favourites, was beautifully simple: a long tarpaulin dragged backwards by a pickup truck, under an ultralight that lifts off at 25 to 30 miles an hour. Even the pilot they hired didn’t think his plane would leave the ground. It flew.

The MythBusters’ test, from the show’s own channel. Nothing loads from YouTube until you press play.

It settled nothing. When a scheduling slip delayed the broadcast, the answering machine at Hyneman’s workshop filled overnight with callers shouting that the show was in on a conspiracy, and on the MythBusters’ own forum plenty of viewers simply refused the result. Looking back in 2025, Savage called it maybe the most controversial story they ever filmed, and the personal backlash against him and Hyneman “spectacular”.MythBusters, episode 97, “Airplane on a Conveyor Belt”, Discovery Channel, 30 January 2008. Savage tells the story behind it in “This Myth Generated the Most Personal Backlash Against Adam Savage and Jamie”, Adam Savage’s Tested, 16 August 2025.

Later that year Randall Munroe, who draws the webcomic xkcd, explained why it never ends. The question hides an ambiguity, he wrote, which people resolve in different ways, and each camp then watches the other make what looks like an elementary mistake in physics. He compared it to Richard Feynman’s lawn-sprinkler puzzle, whose answer strikes everyone as obvious, in opposite directions. That, he wrote, was why the airplane was banned from his forums, along with arguments over whether 0.999… equals 1.R. Munroe, xkcd blog, 9 September 2008, which sorts out the readings with equations.

The Russian Wikipedia kept an article on the puzzle for three years. In 2009 an editor nominated it for deletion, calling it the target of a slow-burning holy war in which every attempt to fix it was buried under amateur edits. The deletion discussion promptly turned into another argument about the plane, and the article was deleted as original research.Википедия:К удалению/21 июля 2009, Russian Wikipedia; the article was deleted on 22 August 2009.

The two readings are the two this page has followed. If the belt matches the plane’s speed, the plane rolls forwards and flies. If it must match the wheels’ own speed, the plane has to be standing still, and the puzzle stops being about a takeoff at all: it becomes a puzzle about tyres, which the belt spins past their rated speed in under half a second.

So most people who say no are answering a fair question. It just isn’t the one about a plane taking off. And once a plane is moving through the air, its wing does the rest. How a wing does that turns out to be a far deeper question, one the experts are still arguing about, and it has a page of its own.