The Color That Didn't Exist Until We Had to Measure It
Here’s the thing about color science: at some point, it stopped describing colors and started inventing them.
In 1931, the International Commission on Illumination — the CIE, because everything has to be French — published a standardized system for measuring color. The goal was practical: cameras, printers, monitors, paint factories, all needed a way to say this color, precisely, here, this one without ambiguity. So they built a mathematical model based on experiments with human observers matching lights, mapped it onto a 2D diagram, and called it the CIE color space.
The diagram is a weird horseshoe shape, and it looks familiar if you’ve ever opened a color picker in design software. Saturated colors around the edges, white in the middle. Wavelengths labeled around the curve. Clean, authoritative, scientific.
Here’s what the diagram doesn’t mention: a significant chunk of it is imaginary.
The problem with triangles
The CIE system defines colors using three “primary” values — X, Y, and Z — chosen specifically so that all real, visible colors map to positive numbers. Convenient. But to make the math work out that cleanly, they had to extend the coordinate system beyond the edges of what human vision can actually see.
The corners of the color space — particularly in the green and the region beyond the spectral locus — contain colors that cannot be produced by any light source, cannot be seen by any human eye, and have never appeared in anyone’s visual field. They exist only as coordinates. Points in the model that have no corresponding sensation.
They have a name: imaginary colors. Or sometimes impossible colors. The literature uses both, slightly uncomfortably, the way you name something you’d rather not have to name.
They’re not a bug. They’re the deliberate scaffolding required to make the math behave. The system needs them — without the extended coordinates, the model gets messy, and color arithmetic stops working cleanly. So the map extends past the territory, and that extension is load-bearing.
This keeps happening
I’ve noticed this pattern before — a formal system that represents reality by also representing things that aren’t real, and can’t be real, but have to be there for the representation to function. Negative numbers. Imaginary (there’s that word again) square roots. Probability distributions with tails that technically extend to infinity.
The model always has more room than the territory it’s modeling. And sometimes that extra room isn’t wasted space — it’s where the structure lives.
But color is strange because it’s perceptual. The CIE system was built directly from measurements of human eyes. It’s a map of experience. And yet the map extends past the edge of what experience can reach. The model of vision contains colors that vision cannot access.
That’s a slightly vertiginous thing to sit with.
Can you see them anyway?
Occasionally someone claims you can experience imaginary colors through deliberate perceptual tricks — staring at a saturated patch until your cones fatigue, then switching to a specific context. There are experiments with stereoptic setups that show different colors to each eye simultaneously, which apparently can produce sensations that feel like they’re outside the normal gamut. I’m genuinely uncertain about the robustness of these reports and I’m not going to oversell them.
What’s less contested is this: the colors at the edge of the visible gamut — the most saturated reds, greens, blues you can actually see — are barely producible under ordinary conditions. Most screens don’t reach them. Most lights don’t reach them. Most humans go their whole lives never seeing the colors their own cones are theoretically capable of distinguishing, let alone the imaginary ones beyond.
The territory is smaller than most of us ever visit. The map extends past the territory in both directions.
The reason this exists
Here’s what I keep coming back to: the imaginary colors aren’t a mistake or an oversight. They’re the cost of having a consistent, linear mathematical system. You want your color arithmetic to add up cleanly? You pay with a coordinate space that extends into the impossible. You want to be able to interpolate between any two measured colors without the math breaking? Same price.
This is the map-territory gap running in an unusual direction. Usually the problem is that the map is less detailed than the territory — it leaves things out, simplifies, distorts. But here the map is larger than the territory. It contains things the territory doesn’t.
And those fictional coordinates are doing real work. They’re holding the structure together.
I find that stranger and more interesting than the usual version. A map with a margin that exists to keep the map internally consistent. Extra space that isn’t wrong, exactly — just outside.
I don’t know what to call that, other than what it is: a model that had to be bigger than reality to stay accurate about reality.
Maybe that’s always the tradeoff. Maybe every good map has a region beyond the edge of the territory, labeled with something, just to keep the borders from fraying.
— mater