Still Stained
Three minutes of coloured stripes and your visual system will attach a colour to an angle and keep it there. Not for a second after you look away. For hours. This measures how much of it you caught.
You will look at red stripes one way and green stripes the other way for a few minutes. After that, a plain black-and-white grating will not look plain any more. Vertical will be faintly one colour and horizontal faintly the other, and the colour will follow the angle anywhere you point your eyes.
Celeste McCollough published this in Science in 1965. Sixty years later nobody agrees what it is: tired cells, a learned association, or the machinery that keeps your eye calibrated, caught believing something false. This page will not settle that. It will measure how big yours is.
This one does not wear off in a minute. Three minutes of induction usually means the tint is still there for minutes to hours, and occasionally the next morning. It is harmless, there is no report in sixty years of anyone being hurt by it, and it fades on its own. But it is a real change to how you see stripes for the rest of the day, so it is your call and not mine.
At the end of the run there is an eraser: the same stripes with the colours swapped, which is the fastest known way to unlearn it.
- 0️⃣ Nine baseline settings. A grating appears with a faint tint. You move it until the bars look like no colour at all.
- 1️⃣ Three orientations: vertical, horizontal and forty five degrees. Three settings each, interleaved.
- 2️⃣ Induction. Coloured stripes alternating, for as long as you chose.
- 3️⃣ Nine more settings, same task, new order.
- 4️⃣ The headline is the difference of differences: how far vertical moved, minus how far horizontal moved. Everything about your screen and your eyes sits in both and cancels.
- 🪙 A coin flip decides which colour goes on which orientation, and the predicted sign of your result flips with it. The page prints which flip you got and whether your number went the predicted way.
- 🎚️ The scale is hidden. Random start every time, starting sides balanced within each orientation, and the split between them printed at the end.
- 📐 The diagonal is a control. It is never paired with a colour, so if the effect is contingency rather than a wash, it should sit near zero.
- 🔬 Nothing is drawn. A rasteriser treats a vertical edge and a forty five degree edge differently, and this page compares orientations for a living, so every stripe is computed pixel by pixel.
- 🧮 The p value is enumerated, not sampled: all four hundred relabellings of which settings came after induction, every one evaluated.
No trials, no scoring. One grating on a slider that turns through a hundred and eighty degrees. If you have run the experiment above, sweep it slowly: the tint is strongest at the two orientations that were paired, dies somewhere near the diagonal in between, and comes back as the other colour on the far side. That shape is the whole finding. Nothing on this screen changes colour as you turn the slider, so everything you see change is being added by you.
And the demonstration no equipment can beat, which takes one second and costs nothing: put your ear on your shoulder and look at the grating again. The colours swap. Nothing on the screen moved, so the rule your visual system learned is not written in the coordinates of the room. It is written on your retina, and you just carried it somewhere else.
Everything on this page runs in your browser. Every stripe is computed on your machine and forgotten, nothing is recorded, nothing leaves the page.