How Dot-Plate Color Vision Tests Work — Confusion Lines and Luminance Noise
Why can some people see the number while others can't? The principle behind pseudoisochromatic plates and how Colorblind Lab computes its plates.
Dot-plate tests are known as pseudoisochromatic plates. Each plate paints a number and its background in two colors that are clearly different to normal color vision but look the same to a particular type of color vision deficiency. The best known are the plates published in 1917 by Shinobu Ishihara of Japan.
Three cones, two signals
Our eyes create color by comparing the responses of the L, M, and S cones. Roughly speaking, “the difference between L and M” signals the red-green direction, and “the difference between S and (L+M)” signals the blue-yellow direction. When L or M is missing or the two become too similar, the red-green signal weakens; when S is missing, the blue-yellow signal weakens.
Confusion lines
If you plot every color on a chromaticity diagram (a map of color with brightness removed), the colors that look identical to a protan dichromat gather on straight lines radiating from a single point. These are confusion lines, and the point where they converge is the copunctal point. Deutans and tritans each have their own copunctal point.
Two colors on the same confusion line are exactly the same color to a dichromat of that type, as long as their brightness matches.
Colorblind Lab uses the same copunctal points as the Cambridge Colour Test on the CIE 1976 u'v' chromaticity diagram — protan (0.678, 0.501), deutan (−1.217, 0.782), tritan (0.257, 0.0). For each plate it picks a center color, then pushes the number color and the background color apart in opposite directions along the line toward that point.
Luminance noise — no reading by brightness
Even two colors on a confusion line will give the number away if their brightness differs. So each dot's brightness is picked at random from six levels. This is the luminance noise that Regan, Reffin, and Mollon (1994) used in the Cambridge Colour Test. To read the number, you have to see the color difference, not the brightness.
Shrinking the color difference step by step
The difference between normal color vision and color weakness is less about “seeing it or not” than about “how small a difference you can still see.” Colorblind Lab presents plates with the u'v' distance between number and background shrinking from 0.100 → 0.019 on the red and green axes and from 0.110 → 0.036 on the blue-yellow axis (in 16-plate mode: 5 plates each for red and green, 4 for blue-yellow).
Calculated with the Machado (2009) model, the color difference between number and background on the easiest red-axis plate (CIELAB chroma difference) is about 54 for normal color vision, but about 1 for the protan dichromat model — effectively gone — while about 15 remains for the deutan dichromat model. Because each axis breaks down for a different type, which axis you miss more often points to your likely type.
| Axis | Type it breaks down for | Color difference steps (u'v') |
|---|---|---|
| Red axis | Protan (protanopia, protanomaly) | 0.100 → 0.019 |
| Green axis | Deutan (deuteranopia, deuteranomaly) | 0.100 → 0.019 |
| Blue-yellow axis | Tritan (tritanopia, tritanomaly) | 0.110 → 0.036 |
| Control plates | None (readable by everyone) | Shown by brightness difference |
How results are read
- On each axis, missing up to one plate is still within the normal range. The hardest plates are close to the threshold even for normal color vision, depending on the screen.
- If two or more are missed, the share of misses sorts the result into mild (half or less), moderate, and strong (85% or more).
- If the miss rates on the red and green axes differ by 30 percentage points or more, the result shows a leaning toward that type; otherwise it stays “hard to distinguish.” Even in clinical testing, dot plates alone are not very accurate at separating protan from deutan.
- Missing control plates lowers confidence in the result, and a repeated pattern of missing easy plates while getting harder ones right is flagged separately.
Limitations
Printed test plates are viewed under standardized lighting and printing, but screen colors differ from device to device. Colorblind Lab computes its colors assuming a standard sRGB monitor, so on wide-gamut or uncalibrated screens the plate colors may drift slightly off the confusion lines. Treat the result only as an estimate for reference, and get an eye exam if you're concerned. Limits of online tests