How our color blind test works
Our test uses pseudoisochromatic plates, the same principle as the Ishihara and HRR tests: a figure made of colored dots hidden in a field of other dots. Instead of copying printed plates, we generate every plate in your browser from published models of human cone sensitivity. This page explains exactly how.
Last updated 2026-11-09.
1. Color science: cone-isolating colors
Human color vision relies on three cone types, sensitive to long (L, “red”), medium (M, “green”) and short (S, “blue”) wavelengths. We convert screen colors (sRGB) into cone excitations using the Smith & Pokorny cone fundamentals in the form given by Viénot, Brettel & Mollon (1999).
For each plate we pick a neutral base color, then make the figure and background differ only in the excitation of one cone type. Two colors that differ only in L-cone excitation lie on a protan confusion line: someone without working L cones receives exactly the same signal from both, so the figure disappears. The same holds for M cones (deutan) and S cones (tritan). People with normal color vision see the figure clearly.
We verify every plate type automatically: each color must sit inside the sRGB gamut, the simulated dichromat must see a color difference of essentially zero (CIELAB ΔE < 0.5), and normal vision must see a clear difference.
2. Luminance noise
Each dot’s brightness is randomly varied by roughly ±20% while keeping its chromaticity. This is the same idea used by modern clinical tests such as the Cambridge Colour Test (Regan, Reffin & Mollon, 1994): it stops people from finding the figure by brightness alone, so only the color signal can reveal it. Dot sizes and positions are produced by a random circle-packing algorithm, so every plate is unique.
3. Plate types
| Plate | What it measures |
|---|---|
| Control | Figure differs in lightness and in all three cone types. Everyone should see it; missing it flags a screen or attention problem. |
| Protan (L-cone) | Hidden from people with missing or altered red-sensitive cones. |
| Deutan (M-cone) | Hidden from people with missing or altered green-sensitive cones. |
| Tritan (S-cone) | Hidden from people with missing or altered blue-sensitive cones. |
| Classification | Two digits: the left one is carried by L cones only, the right one by M cones only. Protans tend to report only the right digit, deutans only the left one. |
| Hidden figure (reverse test) | Strong noise along one cone axis hides a figure that is equally bright to normal vision. Only used on the reverse test. |
4. Contrast levels
Plates on each axis come in several strengths. The numbers below are the cone contrast between figure and background (the figure is shifted up by this amount and the background down by the same amount).
| Axis | Level 1 | Level 2 | Level 3 | Level 4 | Level 5 |
|---|---|---|---|---|---|
| L (protan) | ±13% | ±9% | ±6.2% | ±4.5% | – |
| M (deutan) | ±13% | ±9% | ±6.2% | ±4.5% | – |
| S (tritan) | ±60% | ±42% | ±31% | ±24% | ±18% |
S-cone contrasts are higher because normal sensitivity to S-cone changes is several times lower than to L or M changes. Even the most subtle level is set well above the typical normal threshold, so people with normal color vision should see every plate on a typical phone or laptop screen.
5. Scoring: type and severity
For each cone axis we find the strongest level you failed (a pass rate below 50% at that level):
- Seeing every level gives a score of 0 on that axis.
- Failing only the most subtle level suggests a mild deficiency.
- Failing the middle levels suggests a moderate deficiency.
- Failing even the strongest plates (and the classification plates) suggests a strong deficiency, such as protanopia or deuteranopia.
The axis score is 85% this threshold position and 15% the overall share of missed plates. Severity bands: mild from15, moderate from 40, strong from85 (out of 100).
People with a protan deficiency also see deutan plates a little less clearly (and the other way round), so the type is named only when one red-green axis scores at least 15 points higher than the other. Otherwise we report “red-green, unclassified” and suggest the longer red-green test. “-anomaly” is used for mild and moderate results, “-opia” for strong ones.
Adaptive retesting. If an axis gives an inconsistent pattern (for example a strong plate missed but a more subtle one seen, or a single missed plate), the test adds a fresh confirmation plate at that level, up to3 extra plates per test.
Quality checks. Two control plates appear in every test. Missing half or more of them, or missing nearly everything on all three axes, gives an “inconclusive” result instead of a type.
6. Validation
The scoring is tested automatically with simulated observers: normal vision, dichromats (using the Viénot/Brettel projections) and anomalous trichromats at several severities. Normal observers must score normal on every test mode, dichromats must be typed correctly with strong severity, and moderate anomalous observers must be typed correctly with moderate severity. These checks run on every code change. They verify the logic; they are not a substitute for a clinical validation study with real participants, which we plan to run.
7. Limitations
- Uncalibrated screens. We assume a standard sRGB display. Real screens differ, and features like Night Shift, True Tone, blue-light filters or “vivid” modes change colors. Tritan plates are the most affected.
- Viewing conditions. Lighting, distance, glare and screen brightness all matter.
- Not a diagnosis. Clinical diagnosis uses standardised printed plates under controlled light, arrangement tests and the anomaloscope.
- No achromatopsia detection. The test is not designed to detect complete color blindness or other eye conditions.
- Hidden-figure plates on the reverse test are less reliable than standard plates and are offered for interest only.
8. The simulator
Our color blind simulator uses the Machado, Oliveira & Fernandes (2009) model in linear RGB. Partial severities are blended linearly with normal vision, a common approximation of anomalous trichromacy.
References
- Smith VC, Pokorny J. Spectral sensitivity of the foveal cone photopigments between 400 and 500 nm. Vision Research. 1975;15(2):161–171.
- Brettel H, Viénot F, Mollon JD. Computerized simulation of color appearance for dichromats. JOSA A. 1997;14(10):2647–2655.
- Viénot F, Brettel H, Mollon JD. Digital video colourmaps for checking the legibility of displays by dichromats. Color Research & Application. 1999;24(4):243–252.
- Machado GM, Oliveira MM, Fernandes LAF. A physiologically-based model for simulation of color vision deficiency. IEEE TVCG. 2009;15(6):1291–1298.
- Regan BC, Reffin JP, Mollon JD. Luminance noise and the rapid determination of discrimination ellipses in colour deficiency. Vision Research. 1994;34(10):1279–1299.
- Birch J. Worldwide prevalence of red-green color deficiency. JOSA A. 2012;29(3):313–320.
Ishihara® is a registered trademark of its owner. Our plates are original works generated by software and are not reproductions of the Ishihara plates.