Try it right now: picture a green that is also red — not the brown you get from mixing paints, not the yellow of overlapping stage lights, but a hue carrying both names at once. Almost nobody can hold the image, and that blank is itself a clue. The missing color is not missing from the world; it is missing from the wiring. Human color vision does not record wavelengths one by one like a spectrometer — it compresses them into comparisons, and the comparison that decides red-versus-green can only lean one way at a time. It is the same opponent architecture that explains why magenta can exist even though no rainbow contains it: mixes the system permits are vivid; mixes it forbids simply never get reported.
Two axes, four directions
The opponent account, formalized by Hurvich and Jameson in the 1950s, holds that after the eye's three cone types respond, later stages re-code their signals as differences: one channel signals reddish-or-greenish, another bluish-or-yellowish, and a third carries brightness. A signal can push hard toward red or hard toward green; what it cannot do, in this scheme, is push hard toward both, because they are read off the same dial in opposite directions. Yellow-red, blue-green, and every familiar hue lives somewhere on the permitted map. Reddish green and yellowish blue would require a single dial pointing both ways at once — hence the intuition of impossibility, which is a fact about coding, not about etiquette. That is also why the constraint survives every clever workaround tried in ordinary viewing — you cannot talk a dial into pointing two directions by staring at it harder.
The experiment that claimed to break the rule
In 1983, Hewitt Crane and Thomas Piantanida of IBM reported an attempt to force the issue in the journal Science. Their lever was retinal stabilization: patterns were made to sit perfectly still on the retina, cancelling the tiny eye movements that normally keep vision refreshed. Observers viewed adjoining red and green stripes locked in place this way. Under stabilization, the border between the colors appeared to dissolve, and several observers described something extraordinary — not alternation, not brown, but a single percept that was simultaneously red and green. A companion condition with blue and yellow produced reports of yellowish blue. The authors were careful with their wording, but the paper landed like a dare: perhaps the forbidden colors were merely unseen, not unseeable.
What the follow-ups found
The claim has had a bumpy ride since. A 2001 study by Billock, Gleason, and Tsou re-examined forbidden percepts using retinally stabilized, carefully matched patterns. Their observers did report reddish-green and yellowish-blue percepts under some conditions — but the results depended delicately on stimulus details, and the authors framed their findings as evidence about how the visual system constructs hue from opponent signals rather than as a party trick anyone can learn. Reviewers writing for general audiences have emphasized the same cautions: the percepts are fleeting, they require specialized apparatus, and interpretation remains contested. What is not contested is the everyday conclusion — in ordinary viewing, the red-green dial refuses to point both ways, and no amount of practice at imagining will move it.
Why build a system that bans colors
The opponent design looks like a strange restriction until you consider what it replaces. The retina reports three overlapping brightness signals, not hue names; something must compress that raw output into a compact code the brain can carry at speed. Splitting hue into two difference signals — one for red-versus-green, one for blue-versus-yellow — is exactly such a compression: two numbers plus brightness can locate every color a human can name, using a fraction of the bandwidth a full spectral description would demand. The forbidden mixtures are not a design flaw; they are the price of the compression, the few combinations the compact code cannot express. Every advantage of the system — hue constancy under changing light, vivid contrast between opposites, stable color naming — rides on the same architecture that makes reddish green unreportable. Compression schemes always have an edge like this — a dictionary that cannot spell every word, a format that rounds some values away — and the red-green channel's edge just happens to fall inside experience itself. In this sense the 1983 experiment was probing the compression itself, looking for gaps where the code's shorthand might be pried apart.
Where the internet goes wrong
Search for “impossible colors” and you will find instructions promising to reveal them through staring at colored squares and crossing your eyes. Most of what those demonstrations deliver is ordinary successiveness — the afterimage of one color riding on the real color of another — which flickers or blends but never becomes a stable forbidden hue. The genuine research requires servo-controlled optics that cancel eye motion entirely, a condition no printable JPEG reproduces. The honest takeaway is stranger than the hype: your visual system is a compression algorithm, and reddish green is its rounding error — a combination its own arithmetic cannot express, testable only at the edge of what the equipment, and the eye, can hold still. For a different tour of categories fighting each other, the Color Stroop game shows what happens when a word and its ink color pull the reading system in two directions at once.
reddish green = both ends of one dial ✕ · yellowish blue = both ends of the other ✕