Coding two categories as red versus green leaves them indistinguishable for a meaningful share of viewers

What it is

Encoding two critical categories as red versus green fails to separate them for a substantial share of readers: congenital red-green colour vision deficiency affects on the order of several percent of men and roughly one in two hundred women — far more people than most design teams assume. For these readers, "red" and "green" are not two categories but one neighbouring colour: the opposition between up/down, normal/abnormal, pass/fail collapses entirely.

Why it happens

Red-green discrimination depends on the difference between the L- and M-cone sensitivity curves. The opsins for those cones are encoded on the X chromosome, and variants shift or disable one curve, compressing perceptual colour space along the red-green axis; two colours separated mainly on that axis project to nearly the same perceived colour. Men carry a single X, so one variant is expressed — hence the far higher prevalence in men, while women with two X's need both affected. The red-positive/green-negative conventions and red-green status lights common in charts put the most critical information on exactly the most fragile axis.

Studying it

Clinically, Ishihara plates and the anomaloscope classify type and severity. Visualisation research usually takes one of two routes: simulate the chart through established transformation models (Machado, Brettel-Viénot) and compare task speed and error rates before and after simulation; or run real task-based tests with diagnosed colour-deficient participants. The typical independent variable is palette construction (whether lightness redundancy is present), with category-reading accuracy and time as dependent variables. The cost of cheap simulation is fidelity: models predict a population average and lose individual variation and severity strata, so key conclusions should be re-checked with at least a small panel of real readers.

Where it stops holding

Red-green deficiency is the most common form of colour vision deficiency but not the only one: blue-yellow defects (tritan) are far rarer, severity varies within the red-green group, and not everyone labelled "colour blind" sees the same world. Prevalence figures vary with population ancestry, so textbook percentages cannot be applied to a specific user base unchanged. Simulation is a screening instrument, not a diagnosis: passing simulation does not prove every real reader can decode the chart, but failing it is near-conclusive evidence of a problem.

Applying it

• Audit existing charts: wherever a red-green opposition is the only distinction (gain/loss fills, status dots, heatmap extremes), flag for remediation.
• Attach at least one non-hue channel to critical categories — a lightness step, a shape, a label; red and green may stay, but never as the sole carrier.
• Batch-run palettes through a CVD simulation tool; pairs that visibly merge get replaced or given a lightness step.
• Verification: render the simulated version and ask a colleague who has not seen the original to name each colour's category; any pair they cannot separate is a failing pair.

Related

• Same group: U4.04.2 Lightness differences must guarantee separability · U4.04.3 Must remain readable in greyscale
• Nearby: U4.05.3 Colour should reinforce, never solely carry, information · U4.05.4 Colour-only status cues fail accessibility requirements
• Search terms: colour vision deficiency · protanopia · deuteranopia · colourblind simulation