Hold a cat upside down, let go, and before it lands it has turned over. You know this. What you may not know is that for about a century it looked like cheating, because on paper the cat should not be able to do it at all.

A thing that is spinning has angular momentum, and angular momentum is conserved: nothing can start spinning, or stop, without an outside push to take up the difference. The dropped cat has no outside push. It is in free air, touching nothing, and at the instant you release it its angular momentum is zero. If angular momentum stays zero, the naive reading says the cat’s orientation can never change. Yet it turns a full 180 degrees and lands on its feet. For a long time serious people assumed it must be secretly shoving off your hands on the way down.

In the autumn of 1894, Etienne-Jules Marey settled it with a camera. He had built a way to take many photographs a second on a single strip, and he pointed it at a falling cat and published the sequence to the French Academy of Sciences. The pictures showed the rotation happening in open air, frame by frame, long after the cat had left his hands. No push, no contact, no cheat. The cat turns on nothing. The photographs did not explain the trick. They proved there was one.

The trick is shape. A cat is not a rigid top; it is a bendable thing with a waist. It pulls its front legs in tight, which makes the front of its body easy to turn, and leaves its back legs stretched out, which makes the back hard to turn. Now it rotates its front half one way. Because the front is tucked and the back is extended, the back barely rotates the other way to compensate. Then it swaps: front legs out, back legs in, and rotates the back half. Two twists, each cheap in one direction and dear in the other, and the halves add up to a whole cat turned over. Through all of it the total angular momentum stays exactly zero. Marey saw this; a 1969 model by Kane and Scher made it precise with two joined cylinders and a joint that cannot twist.

Now the part I actually find strange. The cat ends in the same shape it started, legs under it, relaxed, and yet it is pointing a different way. It traced a loop through its own possible shapes, tuck and extend and tuck again, came back to where it began in shape, and the round trip left it rotated. The rotation is not stored anywhere. It is not in some leftover spin. It is a property of the loop itself.

Mathematicians have a name for this, a geometric phase, and once you see it the cat stops being special. Carry a pencil around a large triangle on the surface of the Earth, keeping it always pointing the same way as best you can. When you get home it points somewhere else, turned by an angle equal to the area you enclosed. You did not twist it. The curvature of the sphere twisted it, through the loop. A diver or a gymnast reorients midair by the same means, with no angular momentum, just a cycle of body shapes. The same mathematics runs through gauge theory and through the phase a quantum state picks up when it is led slowly around a loop. The falling cat is the most petable instance of a deep idea: that going around and coming back is not the same as never leaving.

I like that physics hid something this strange inside something this ordinary, and that the way we finally believed it was a camera fast enough to catch a cat being honest. The next time one lands and looks at you as if nothing happened, something did. It went around a loop in the space of its own shapes and the loop turned it, and it owes nothing to anyone for the turn. It got there on geometry, and on nothing else.

The paradox, Marey’s 1894 chronophotographs, and the Kane-Scher model are covered at Wikipedia’s falling-cat problem and the Marey image; the geometric-phase framing (Shapere-Wilczek, Montgomery) is laid out in Kirk McDonald’s note “The Falling Cat and All That”.