How to Think About More Than Three Dimensions
How to Think About More Than Three Dimensions
You can’t picture four dimensions. Nobody can — not mathematicians, not me. But you don’t need to picture a dimension to read one, and you’ve been doing the hard part your whole life. Look at any photograph: it’s flat, two dimensions, and yet you rebuild the room inside it instantly — the depth that isn’t on the paper. (You also live inside a fourth dimension you rarely think about: time, which you move through one now at a time.) So “more than three” isn’t a wall. It’s something your mind already handles. Let me show you the one trick the whole Atlas leans on, and the constellation will stop looking impossible.
The shape is in the motion
Start with what a flat picture really is: a shadow. The world has more directions than the page, so the page keeps a couple and throws the rest away. That’s honest, but it’s lossy — a great many different things can cast the very same flat shadow, which is why a single still picture can leave you guessing what made it.
Here is the trick that gets the lost dimension back: motion.
A cloud of points that genuinely lives in five dimensions, shown as a slowly-tumbling 2-D shadow with a pause button. Frozen, it’s a smear; you can’t tell what’s there. Turning, the structure swims up — crowds part and gather, a shape you can almost hold. Pause it, and the shape dissolves back into smear. It was never in any single picture. It lives in the motion, and your eye assembles it from the turning. That is how you come to “see” a dimension your screen can’t hold: not in one frame, but across many.
Which shadow do we show you?
Motion rebuilds what a flat shadow loses. One question is left — the one the Atlas itself had to answer. If a thing has thousands of directions and your screen has two, which two do you point it at? Turn it to a random angle and you mostly get noise.
A streaky cloud hides in seventeen dimensions: two directions carry real structure, the other fifteen are just noise. Let it tour, then press PCA — it spins to the plane where the action lives and sets the fifteen noise directions aside, but the two streaks sit at an angle, because PCA doesn’t care which way the plane is turned. Press PCA + varimax and the plane turns until each streak lands on a bearing: one running right, one running up. Now each axis is a single thing — something a reader can name. That’s the whole division of labor: PCA finds where to look; varimax turns it until the axes mean something. (Old mathematics, both — Hotelling, 1933 and Kaiser, 1958; the primer tells the fuller story.)
So that’s the Atlas
Now you know what you’re looking at. Every dot is a real document, living at its address in a four-thousand-direction space. The constellation on the page is a flat shadow of that space — aimed by PCA at the angle that reveals the most, squared up by varimax so the bearings have names. And when it tumbles, it is doing exactly what the cloud above did: handing you, through motion, a dimension your screen could never hold still.
What you can’t see all at once, you can watch over time. That’s the whole reason a map can take a four-thousand-direction shape, turn it onto a flat screen, and let you almost see it.
Now go read one — How to Read the Atlas.