About This Site

A free color blind test that generates its own Ishihara plates, checks them against a model of color blind vision before showing them, and never sends your answers anywhere.

Why This Site Exists

Most online color blind tests are scans of the printed Ishihara series, passed through a JPEG encoder and an unknown color profile before they reach your screen. By the time a scan has been resized, recompressed and served, the colors have moved. The plate still looks like an Ishihara plate, but the precise color relationship the test depends on has been lost, and nobody checks whether it survived.

We wanted a test where that relationship is guaranteed by construction rather than by luck, and where the result explains itself instead of just announcing a number.

How the Plates Are Made

Color vision works by comparing the responses of three cone types in the retina: L for long wavelengths, M for medium and S for short. Color blindness is the loss or mistuning of one of them.

For every plate, the figure color and the background color are computed in LMS cone space and separated along the exact axis the target deficiency cannot see. Someone with normal color vision sees two distinct colors. Someone missing that cone receives an identical signal from both, so the number disappears into the background. This is done live in your browser each time you take the test.

Every plate is then rendered through a dichromat simulation and measured. If the figure has not actually collapsed into the background, the plate is rejected and regenerated. The measured color distance between figure and background under simulation sits between 0.5 and 1.1, against 4.4 to 5.0 for the tabulated approach we tried first. You can read the mechanism in detail on the Ishihara test page.

What That Buys You

• The colors are correct by construction, not by hoping a scan survived a round trip through image compression.
• Every visitor gets a different arrangement, so the answers cannot be memorised, looked up or shared.
• Nothing copyrighted is reproduced. The printed 1917 series remains the property of its publisher.
• Your answers stay on your device. There is no account, no email and no server that could store a result.

Where the Numbers Come From

The simulation of color blind vision follows the method published by Vienot, Brettel and Mollon in 1999, which is the standard approach used in accessibility tooling. Prevalence figures quoted on the types of color blindness page describe populations of Northern European descent, because that is the population most published surveys sampled, and the page says so rather than presenting those rates as universal.

The design of the plate classes, the three second viewing rule and the illumination conditions all come from how the printed test is actually administered.

What We Do Not Claim

This is a screening tool. It cannot grade severity, it cannot reliably separate a missing cone from a mistuned one, and it is not accepted as evidence for a job, a licence or a clinical record. Your display is not calibrated, your room is not measured, and neither of those is true of a printed plate under a daylight lamp.

We are not clinicians and we do not present ourselves as such. Where the honest answer is see an optometrist, the site says so. The disclaimer sets this out in full.

Accessibility

A site about color vision that fails its own test would be indefensible, so the interface carries no red or green signalling of its own. Buttons are dark on light, the only accent is blue, and every state is carried by shape, position or text as well as color. The site is keyboard operable, the test can be answered with the number keys, and progress is announced to screen readers.

Contact

Corrections and questions are welcome, particularly from eye care professionals. Use the contact page.