Why some colors cannot exist — and how they can.

The opponent ledger

Human color vision is not a palette of independent paints. After the three cone types (L, M, S — loosely, long, medium, and short wavelength) have fired, the retina and cortex recode them into opponent channels: red versus green, blue versus yellow, and black versus white.

That recoding is why you never see reddish-green or yellowish-blue in the world. More red is, by construction, less green. The channel cannot read both directions at once, any more than a number can be both positive and negative. Mix red and green light and you do not get a forbidden hue — you get yellow, or brown, or gray. The mixture is a third color. The binary is not allowed.

Chimerical color

Afterimages punch a hole in that ledger. Stare at a hue long enough and the cones that report it adapt: they fire less. Look away, and the silent cones are counted as a vote for the opponent. A red field leaves a cyan ghost. The ghost is not on the wall. It is a deficit in the report.

Philosopher and neuroscientist Paul Churchland called the extremes of this trick chimerical colors — hues that no physical light can match:

• Stygian — an afterimage of blue thrown onto black. Blue has a hue, but black is already zero luminance, so the blue is seen as darker than black.
• Self-luminous — an afterimage of red thrown onto white. The red has nowhere to go above maximum white, so it is seen as glowing, fluorescent, emitting.
• Hyperbolic — an afterimage added to a saturated hue of the same family. The result is denser than any physical sample of that hue: more orange than orange.

Impossible colors

In 1983, Hewitt Crane and Thomas Piantanida published a brief report in Science: with a retinal image held still, observers could see reddish-green and yellowish-blue — colors that “should not” occur. The method was filling-in. Two fields, a thin gap, a gaze that does not wander. The cortex interpolates the seam, and sometimes the interpolation is a color with no name.

Later work is mixed. Some observers report a genuine binary hue. Others report rivalry, or a muddy olive, or nothing. Without optical stabilization the effect is fragile. That is not a failure of the chamber. It is the point: the forbidden color is a state of the observer, not a wavelength.

Olo

In April 2025, James Fong, Ren Ng, Austin Roorda and colleagues described a different kind of new color. Using a system they call Oz, they tracked the eye and painted laser light onto individual photoreceptors in living human retinae — thousands of them, targeting only M-cones, leaving L and S unstimulated.

No natural light can do that. Any teal in the world hits M-cones together with their neighbors. The isolated M signal was reported as a blue-green of unprecedented saturation. They named it olo.

A screen cannot isolate a cone class. The Olo chamber in this laboratory is an analog, not a replica: it fatigues L and S until a teal arrives with almost no opposition. Some people describe a color they have no name for. Some describe ordinary mint. Both reports are data.

What this cannot do

These experiments do not put new photons in the room. They do not make you tetrachromatic. They will not match the laser-isolated olo of the Oz system. They will not work well on a dim phone in daylight.

They will, if the conditions are right, briefly show you a color that has no swatch — a color whose coordinates do not exist in RGB, in paint, or in the sky. That is as close as a page of light can come to showing you something it cannot display.