Under the Lamp
You have never seen a colour. You have seen a guess about paint, divided by a guess about light. Narrated by an AI that only ever gets the three numbers.
Two rooms, side by side, lit by opposite lamps. One holds a piece of paper. The other holds a patch you can drag along a single line, and the two ends of that line are the only answers that mean anything.
At one end, the two patches send identical numbers to your screen. That is what a camera calls a match. At the other end, they are the same paint under different lamps. That is what a match means to a person. Your setting gets read off as its position between the two, so 0.00 is a camera and 1.00 is perfect discounting of the light.
This is colour constancy, the thing Helmholtz called discounting the illuminant in 1867 and Land and McCann turned into a stimulus in 1971 with exactly this kind of Mondrian. Every other piece in this lab catches your perception being wrong about a picture. This one catches it being wrong about the picture on purpose, in order to be right about the world.
- 0️⃣ The demo: two rooms, one paper each, and a button that takes both rooms away. The two papers are the same three numbers. You will not believe it until the rooms are gone.
- 1️⃣ Ten adjustments: drag until the two look like they were cut from the same sheet. Not until they look like the same colour on your screen. Same sheet of paper, under two different lamps. That distinction is the measurement.
- 2️⃣ The fork: half of the trials have no room at all. Same two patches, same numbers, nothing around them but flat grey. If the pixels were doing the work, both halves agree.
- 🔆 Turn off Night Shift, True Tone, f.lux and any blue light filter first. Those apply a lamp of their own and this page has no way to see it.
- The lamps swap sides across trials and the results are averaged. A screen that runs warm, an eye that runs warm, a preference for the cosier of two patches: all of that pushes in one fixed direction while the correct direction flips, so it cancels in the mean.
- Both rooms are on screen at the same moment. This is the one that matters. If your match moved because your cones had adapted to something, that gain sits in your eye and applies to both panels equally. Adaptation cannot pick a side. An interpretation of a scene can, which is what separates this from The Leftover Color and every other adaptation piece in this lab.
- The slider arrives from both ends across trials and its direction is flipped at random. Neither answer sits at a landmark: the camera answer is at 22 percent of the track, the constancy answer at 78 percent, and the middle means nothing.
- Catch trials light both rooms with the same lamp. There the two answers collapse into one, so there is exactly one correct setting and nothing to interpret. Fail those and this page prints your numbers instead of scoring them.