The Occasional Jam No. 31

No. 31 soapbox

Colors That Don't Make Sense

8 min read

Try to picture a reddish green. Not brown, not olive, not red polka dots on green. One color that's both at once, the way orange is both red and yellow.

You can't, and it isn't a lack of imagination. Nobody can. I fell into this after seeing a post about hyperbolic orange, a color that doesn't really exist but that you can see anyway, if you stare at cyan for half a minute first. The further I read, the more all of it came down to one idea.

A color isn't something light has. It's a report your eyes file about light, and the report has a format.

Some colors don't fit the format, so nobody has ever seen them. Some can be forced into it for a few seconds, if you know the trick. Others fit fine but won't sit still long enough for a screen or a camera to hold them. This post has a toy for most of them.

Why there's no reddish green

Your eyes sense color with three kinds of cells called cones. Each kind answers to a broad band of light. One leans toward long, reddish wavelengths, one toward medium, greenish ones, and one toward short, bluish ones. The bands overlap a lot, so almost any light sets off at least two kinds at once.

Your brain never gets those three readings, though. The wiring right behind the eye compares them and sends on the comparisons: how red against how green, and how yellow against how blue, plus a third for light against dark. Each comparison works like a seesaw. It can tip one way, tip the other, or sit level. It can't tip both ways at once.

That's the whole reason there's no reddish green. Red and green sit at the two ends of one seesaw, and yellow and blue at the ends of the other. Orange is fine, because red and yellow live on different seesaws and each can lean its own way. Drag the light below across the rainbow and watch both seesaws move.

Three cones, two seesaws A simplified model

S, shortM, mediumL, long
Green Red
Blue Yellow

Red-green tips hard toward red, and blue-yellow tips toward yellow. You'd see orange.

The curves and both seesaws come from the standard 1931 color-matching data, and the seesaws sit level right where people see pure blue, green, and yellow. The swatch paints what the two seesaws report, which is the color you'd see.

Now tap Red + green. You'd expect reddish green. You get yellow, because the two lights push opposite ends of the same seesaw and it settles level, while the other one still leans yellow. There's no box for reddish green, so the answer comes out as something else. Blue + yellow does the same to the other seesaw and comes out white. Your screen leans on this every second it's on.

Red + blue shows the opposite puzzle. No single wavelength of light is magenta, which is why it isn't in the rainbow. Your eyes file it anyway, red on one seesaw and blue on the other, a perfectly good reading that just takes two lights to produce.

For the builders: where the curves and seesaws come from

The three curves are the long, medium, and short cones, worked out from the standard 1931 color-matching functions. The seesaws are Hurvich and Jameson's 1957 approximation of the opponent channels: red-green is the x curve minus the y curve, and yellow-blue is four tenths of y minus z. Nothing in either was tuned by hand. The color names in the readout are my own rough bins.

Those two simple differences come out level near 478, 499, and 578 nanometers, which is where people report pure blue, pure green, and pure yellow. The idea is older than the cells it describes. Ewald Hering proposed opponent colors in the late 1800s, and neurons that compare cones this way didn't turn up until the 1950s.

Impossible, mostly

So reddish green should be off the table for good. In 1983 two researchers went after it anyway. They showed people red and green stripes and used an eye tracker to pin the stripes in place on the retina, so the edges between them stayed put even as the eye moved.

Staring at this won't do it, because your eyes never stop making tiny movements and that keeps the edges alive. Drain the color and you can see why these two were picked: they're the same brightness.

A border held perfectly still on the retina fades away, and the brain fills in what's left. Once the edges faded, some of the people saw the two colors flood into each other: a reddish green, or a color they couldn't name at all. Yellow and blue stripes did the same thing.

Not everyone buys it. A 2001 study only got the effect when the two colors were equally bright, which is why the stripes above go to one flat gray when you drain them. A 2006 study argued people were just seeing an ordinary in-between color. It's still argued about.

Colors past the end of the seesaw

There's a second way to break the format, and anyone can do this one. A seesaw's level point isn't fixed. It drifts to match whatever's around, which is how the same paint can look like two different colors.

One color, printed twice. On black it leans orange, and on white it's plainly brown. Connect them and the bar seems to change color as it crosses over. It doesn't.

Brown is orange read against something brighter, and no light is brown on its own. That's the level point drifting across space.

It drifts across time, too. Stare at one color long enough and the cells reporting it tire out, so the level point slides away from that color. Look at something else, and for a few seconds the seesaw reads as if it were pushed past its end, further than any real light could push it.

The philosopher Paul Churchland sorted the colors you can see this way into three kinds, and named them. No light can make any of them.

  • Hyperbolic More saturated than any light can be. Cyan, then orange.
  • Stygian As dark as black, and still a color. Yellow on black, then black.
  • Self-luminous Glows brighter than the white around it. Green on white, then white.

Each toy below runs one recipe. Press start, keep your eyes on the dot for 30 seconds, and keep them there when the field changes. Full screen works better, since a bigger field makes a bigger ghost. Any eye movement at all smears it.

Hyperbolic orange is the one that started this. Cyan works the cells that orange barely touches. Tire them out, and orange arrives with nothing pushing back, so the seesaw tips further toward orange than orange light ever could.

Hyperbolic orange

Stare at the left dot for 30 seconds without moving your eyes, then jump to the right dot.

After the flip the whole field is one flat orange, as pure as a plain web color gets. The disc is your eyes' doing.

Stygian blue is named after the river Styx. Yellow tires the yellow end of one seesaw, and its brightness tires the light-dark one too. Then the field goes black, the darkest thing your screen can show, and the ghost manages to be blue and darker still.

Stygian blue

Stare at the left dot for 30 seconds without moving your eyes, then jump to the right dot.

Yellow on black, then nothing but black. The screen is showing plain black where the blue appears.

Self-luminous red runs the other way. The disc is a darkish green on bright white, so the patch of your eye under it gets a break from the glare. When the field goes white, that patch is more sensitive than the rest, and its red ghost seems to glow brighter than the white around it.

Self-luminous red

Stare at the left dot for 30 seconds without moving your eyes, then jump to the right dot.

Green on white, then plain white. The page is the same white everywhere. The glow is the break your eye got.

Make your own

Any pair works, though some work much better than others. The ghost is roughly the opposite of what you stared at, so the strongest pairs stare at one color and then look at its opposite. Tap a pair to load it, pick your own, or swap the two around.

Hyperbolic green

Stare at the left dot for 30 seconds without moving your eyes, then jump to the right dot.

The disc sits on a gray matched to its brightness, so only the color does the work.

Colors a screen can't hold

Every color in this post so far came out of your screen, and your screen can only make three: a red, a green, and a blue, in dots too small to see. Everything else is the seesaw trick from the top of this post, run on purpose. Drag the zoom down to actual size and watch the stripes turn into the color.

Under the glass ×32

No yellow light anywhere. Red and green dots, side by side, too small to tell apart.

Drawn as the classic stripe layout. Many phones arrange their dots in other patterns, but nearly all of them mix red, green, and blue the same way.

The yellow on your screen is red and green dots pulling opposite ends of the red-green seesaw until it sits level, while the blue-yellow one leans yellow. A pure yellow wavelength does the same thing to both seesaws, so your eyes can't tell the two lights apart, and as far as you're concerned they're one color. White is the same trick with blue added.

If you want to see the real dots, put a small drop of water on your screen and look through it. It works as a magnifying glass.

Three lights can't reach every color, though. Here's the map of everything light can make for human eyes, with the screen's share drawn in.

Every color light can make for human eyes fits inside the horseshoe. A standard web color can only reach the bright triangle, so the faded rest is a stand-in. The two marked outside the edge are placed roughly.

The curved edge is the rainbow, one pure wavelength at a time, and most of it sits outside the triangle. That's why a rainbow, the stripe from a prism, or a green laser pointer looks more intense in person than in any photo of it. The photo has to pull those colors in to the triangle's edge.

In 2025 a team at UC Berkeley went past the horseshoe itself. Its edge is there because no light can reach the green cones alone; the long cones always come along. So they mapped a patch of each subject's retina cone by cone and aimed tiny doses of laser light at the green cones only. Five people saw the result, called it a blue-green of unprecedented saturation, and named it olo.

That last part was measured, not just described. Asked to match olo against nearly pure laser light, every subject had to wash olo out with white before the two would match. Not everyone agrees it's a new color rather than an extreme green. Either way, it's off the map, where no screen reaches.

Colors that don't stay

Some colors are easy to make and hard to keep. Shine your phone's flashlight through a fingertip in a dark room. The glow is a deep, warm red, and it looks lit from inside because it is. Blood soaks up blue and green light and lets the red through, so what comes out the other side is only the red that survived the trip.

Photos rarely do it justice. In person the glow is far brighter than the dark around it, and a photo has to squeeze that gap down to what a screen can show. It's a self-luminous color for real, no staring required, and it lasts exactly as long as the flashlight does.

The strangest one on my list needs no light at all. In 1896, months after X-rays were discovered, a physicist named Brandes sat in the dark until his eyes adjusted, put his eye up to an X-ray tube, and saw a faint blue-gray glow that seemed to come from inside the eye itself. People getting radiation treatment near the eye still report blue flashes today. Please don't go looking for this one.

So when a post says a color doesn't exist, it's worth asking which kind it means. Reddish green really doesn't. Hyperbolic orange does, for about five seconds, if you hold still.

Sources

  1. Churchland, Chimerical colors: some phenomenological predictions from cognitive neuroscience, Philosophical Psychology 18(5), 2005 Where the three kinds get their names: hyperbolic, stygian, and self-luminous.
  2. Hurvich and Jameson, An opponent-process theory of color vision, Psychological Review 64, 1957 The two seesaws in the first figure.
  3. Wyman, Sloan, and Shirley, Simple analytic approximations to the CIE XYZ color matching functions, JCGT 2(2), 2013 The color-matching curves the cones and the seesaws are drawn from.
  4. Crane and Piantanida, On seeing reddish green and yellowish blue, Science 221, 1983 The stabilized stripes.
  5. Billock, Gleason, and Tsou, Perception of forbidden colors in retinally stabilized equiluminant images, JOSA A 18(10), 2001 The replication that found the colors only at equal brightness.
  6. Hsieh and Tse, Illusory color mixing upon perceptual fading and filling-in does not result in forbidden colors, Vision Research 46, 2006 The case that people saw an in-between color instead.
  7. Fong et al., Novel color via stimulation of individual photoreceptors at population scale, Science Advances 11(16), 2025 Olo: five subjects, the blue-green description, and the match that needed white added to olo.
  8. Olo (color), Wikipedia Where the doubts about calling olo a new color are collected.
  9. Bloomfield, flashlight through a hand, How Everything Works Why the light that gets through a hand comes out red.
  10. Wilhelm Röntgen and the Invisible Light, ORAU Museum of Radiation and Radioactivity Brandes, the dark-adapted eye, and the blue-gray glow.
  11. Mathis et al., Mechanisms of phosphenes in irradiated patients, Oncotarget 8(38), 2017 The blue glow and flashes reported under radiation, then and now.
  12. Metamerism (color), Wikipedia Why a screen's red and green dots and a pure yellow wavelength look like one color.
  13. CIE 1931 color space, Wikipedia The horseshoe map, its rainbow edge, and the line of purples.
  14. Impossible color, Wikipedia The stare-then-look recipe for each of the three kinds.
  15. Opponent process, Wikipedia Hering's opponent colors, and when the neurons turned up.