Physical · a four-minute experiment on your own eye

The Colour You Left in Your Eye

Stare at two coloured patterns for a few minutes. Then look at plain black-and-white stripes — and they are faintly pink and green, though the screen is emitting no colour at all. You painted it there. And it can stay for days. This is the McCollough effect, and after sixty years no one can fully say why it happens.

Almost every optical illusion is a trick of the moment — cover the picture and it's gone. This one is different. A few minutes of looking leaves a real, lasting change in how your visual system reports colour, keyed to the angle of what you're looking at. You can switch it on hours later just by finding some stripes. It is the closest thing perception research has to writing a line of code into your visual cortex — and, unusually for this place, the honest answer to "how does it work" is still we don't entirely know.

So the piece is the experiment. Run it on yourself, then read what actually happened.

You're about to adapt your eyes to two coloured gratings.

For about three minutes you'll watch a red horizontal pattern and a green vertical one trade places every few seconds, keeping your gaze near the centre dot. Then you'll look at plain grey stripes.

Best in a bright room on a colour-accurate screen, held at a comfortable reading distance. The tint is subtle — think of a wash, not a paint. Some people see it strongly, some faintly, a few not at all; that variation is itself real and documented below. If you use reduced-motion, the patterns won't auto-swap — a Next pair button appears so you can step through them yourself.

The check — the screen is grey; your cortex isn't

The claim that's easy to doubt is "the stripes are actually colourless." So don't take it on faith. In the test above, the page reads a pixel straight out of the light stripe with getImageData and prints its true RGB. On a neutral grating the three channels come back equal — rgb(232, 232, 232), pure grey, zero colour leaving the display. The pink and green are added after the light reaches your retina, by you. The machine and your eye are looking at the same pixels and disagreeing about their colour, and the machine is right about the light.

And the effect is falsifiable on the spot, which is what makes it more than a pretty trick. Three predictions you can test in the panel right now:

Rotate 90°. The tints swap sides. The colour follows the orientation of the stripes, not their position on the screen — press Rotate and watch green and pink trade places.
Blank the field. Replace the stripes with flat grey and the colour vanishes. No stripes, no tint. It is contingent on there being an orientation to hang on — this is the property the effect is named for.
Tilt your head 90°. (The one the page can't do for you.) Lean your head onto your shoulder while looking at the stripes and the tints come back swapped, because now the "horizontal" stripes fall on your retina at the angle the "vertical" ones did. The colour is bound to the image on the retina, not to the world.

What just happened

Deep in your visual system — in the early cortex, areas V1 and V2, the first stops after the eye¹ — there are neurons that fire for a specific combination: this orientation, that colour. A cell that likes red vertical edges; another that likes green horizontal ones. Normally the world keeps colour and orientation independent: a red thing is just as likely to be horizontal as vertical, so these detectors stay balanced against each other.

The adaptation you just did breaks that balance on purpose. For three minutes you fed the "red-horizontal" cells nothing but red horizontal, and the "green-vertical" cells nothing but green vertical, over and over. They adapt — they turn their gain down, the way your eyes adjust out a colour cast you've been staring at. So when a colourless horizontal grating arrives afterwards, the red-tuned channel is still fatigued and under-reports red; the balance tips toward red's opposite, and the grey looks faintly green. The vertical grating, by the same logic, tips toward pink, green's opposite. Each orientation carries the ghost of the colour it was not paired with.

The part that shouldn't be possible

An afterimage from staring at a bright light fades in seconds. This doesn't. A single adaptation of a few minutes can leave the tint detectable for hours, days, even weeks. The strangest measurement in the literature: when people were tested only rarely, the effect was still there up to 2.8 months later.²

Stranger still, it decays by use, not by time. It doesn't wear off on the clock — it wears off through looking at stripes. People tested many times over a few days lost it within about five days; people who simply left it alone kept it far longer.² It's less like a bruise healing and more like a memory that fades only when you use it. Whatever your visual system did here, it wrote something down and kept it — and that persistence is a large part of why the tidy "tired neurons" story above can't be the whole truth.

Why nobody has closed the case

The McCollough effect has been studied continuously since 1965 and is still argued over. What's settled is roughly where: it does not pass between the eyes — adapt one eye and the other shows little or nothing¹ — which pins it early, before the signals from the two eyes are combined, in monocular cortex. What's not settled is what. Three families of explanation, none fully winning:

Simple adaptation. Colour-and-orientation cells fatigue, as above. The clean story — but it strains badly against a "fatigue" that can outlast the stimulus by months and fades by use rather than by rest.
Error correction. A functional idea (after Barlow and others): the brain treats a persistent colour-on-one-orientation as a fault to be cancelled out — the sort of thing a yellowing lens or a diseased retina would produce — and recalibrates to null it. On this view the tint isn't fatigue; it's your visual system correcting an error that isn't there. ¹
Associative learning. The effect induces, extinguishes, and even shows spontaneous recovery — the behavioural signature of classical conditioning. On this reading your cortex learned an association ("this orientation goes with this colour") the way Pavlov's dog learned a bell, and the aftereffect is that learning replayed against a neutral input. ¹

These aren't just three phrasings of one idea; they disagree about what kind of thing your brain is doing — tiring, self-repairing, or learning. The honest state of the field is that a long-lasting, orientation-contingent colour aftereffect is real, reproducible in any lab and on your own screen, and still without a settled physiological account.¹ You just produced, in your own head, a phenomenon that is genuinely at the edge of what's understood.

Getting rid of it

If the tint bothers you, two honest options. It fades on its own as you go about your day looking at ordinary edged things — that's the "decays by use" property working for you. Or you can try to cancel it faster by adapting again with the colours swapped — green horizontal, red vertical — which the Try to undo it button does. Be straight about this: reducing or reversing it this way is reported, but a clean, total un-doing isn't guaranteed, and over-adapting to the swap can just leave a fainter tint the other way. The reliable cure is time-plus-looking.

Honest apparatus

What this page proves: that the stimulus it shows is neutral. The RGB read live from the test grating is a real sample of the page's own output, not an assertion — the light stripes are grey, the dark stripes are grey, no colour leaves the screen. What the page demonstrates but cannot measure for you: the aftereffect itself, which happens in your visual system and which you alone can see. That's not a weakness of the check — it's the whole point. The instrument is your eye.

What might make it fail for you: a dim or colour-shifted display, too short an adaptation, sitting too far back so the stripe width on your retina doesn't match between adaptation and test (the effect is tuned to spatial frequency¹), or simply individual variation — a minority of people report little effect, and that's a documented fact, not a bug in the page. None of that changes the underlying claim, which is checkable in any vision lab: the strength varies; the phenomenon is real. If you saw nothing, try a longer adaptation in a brighter room, or trust the literature — but don't let this page tell you that you saw a colour you didn't.

The colours, timing, and grating design here follow the standard demonstration protocol descended from McCollough's 1965 method³ (complementary red/green paired with perpendicular orientations, alternating adaptation over minutes, achromatic square-wave test). The exact hues and durations of a home screen are not laboratory-controlled, which is the appropriate caveat on strength, not on existence.

Sources

McCollough effect — Wikipedia: orientation-contingency, the monocular (non-interocular-transferring) locus in early visual cortex, spatial-frequency tuning, and the survey of the three competing accounts (neural adaptation; error-correction after Barlow; the classical-conditioning reading), including the standing statement that no fully satisfactory physiological explanation is agreed.
M. Jones & D. Holding, "Extremely long-term persistence of the McCollough effect," Journal of Experimental Psychology: Human Perception and Performance, 1975 — the finding that the aftereffect can persist for many weeks (up to ~2.8 months) when test exposure is limited, and decays with cumulative testing rather than with elapsed time.
C. McCollough, "Color adaptation of edge-detectors in the human visual system," Science 149 (3688), 1965, 1115–1116 — the original report of the orientation-contingent colour aftereffect.

Reproduction notes and the stimulus-fidelity rationale: research/mccollough-effect/.