Color Vision Challenge

Your score is a ΔE — the standard measure of how different two colours look. It is computed in CIELAB, a colour space built so that equal distances correspond to roughly equal perceived differences. That property is the whole point: it lets the test make a red pair and a blue pair equally hard, which RGB cannot do.

A lower score is better. It means you spotted a smaller difference.

The test uses a three-lives ladder. Every level shaves about 13% off the colour gap, so the difficulty climbs geometrically. You keep going until three mistakes end the run, and the level you reach is the score.

Ending on a guess is normal and expected — the ladder is built to outrun everyone eventually. Where it outruns you is the interesting part.

The early levels are deliberately easy. They are there to get your eyes and your scanning habit warmed up before the gap gets genuinely small, which is usually somewhere around level 12–16.

The two measure different things and it is worth keeping them apart:

• This test measures sensitivity — how small a difference you can resolve, in every hue direction. A trichromat with tired eyes and a dim screen scores poorly here without having any deficiency at all.
• A plate test looks for a missing axis — colours that collapse together specifically because a cone type is absent or shifted. It is diagnostic in a way a threshold number is not.

If you scored badly here, that alone means little. If you scored badly and your errors piled up on one axis, that pattern is worth following up with a proper colour blindness screening and, if it persists, an optometrist.

A threshold measured through a screen is a measurement of you and the screen. The usual suspects, roughly in order of how much damage they do:

• Night mode or a blue light filter. Warms everything and compresses exactly the differences being tested. This one is worth several ΔE on its own.
• Low brightness. Colour discrimination falls off sharply as luminance drops — the cones need light to work with.
• Panel type. A cheap 6-bit panel dithers between levels it cannot display, which can make near-identical colours flicker into visibility, or vanish entirely.
• Ambient light. Sunlight on the screen washes out low contrast differences; a very dark room shifts adaptation the other way.
• Age. The lens yellows gradually through adult life, which costs sensitivity in the blue direction first. It is normal and gradual.

Because of all this, the useful comparison is you against yourself — same device, same lighting, a week apart. A score from a phone in the sun and one from a calibrated monitor are not the same measurement.

Partly, and not in the way people hope. The cones you have are the cones you have — no amount of practice adds sensitivity to the receptors. What does improve is everything downstream: knowing what to look for, scanning systematically instead of randomly, and not being fooled by the surrounding squares.

That improvement is real but it plateaus quickly, usually within a handful of runs, and it is task-specific. Getting better at this grid does not transfer to matching paint in a shop. Expect a modest gain on your first few attempts and then a stable number — the stable number is the interesting one.

If your score drops noticeably over months on the same device, that is a different signal. Colour discrimination can decline with cataract, with some medications, and with conditions affecting the optic nerve or retina. A gradual change across both eyes usually reflects normal ageing of the lens; a change in one eye is worth mentioning to an optometrist.