Colour-Blindness & Inclusive Colour

'Colour blindness' is a misleading name - almost no one sees the world in grey. The accurate term is colour-vision deficiency (CVD): the eye's colour channels are shifted or reduced, so certain colours become hard to tell apart. It is common enough that every designer is designing for it whether they know it or not - roughly 1 in 12 men (about 8 percent) and 1 in 200 women have some form, because the most common types are carried on the X chromosome.

Recall the three cone types from Module 0 - long (L, reddish), medium (M, greenish) and short (S, bluish) wavelength. CVD is a story about those cones:

Type         cone affected   confuses         prevalence
-----------------------------------------------------------
Deuteran     M (green)       red <-> green    most common
Protan       L (red)         red <-> green    common
Tritan       S (blue)        blue <-> yellow  rare
Monochromacy  ~no colour      almost all       very rare

Red-green deficiencies (deuteran and protan together) are by far the most common, which is why the classic accessibility failure is a red/green pairing - the ticked-green, crossed-red form; the red-and-green status light; the metro map that separates two lines by red versus green alone. Blue-yellow (tritan) deficiency is much rarer and affects men and women about equally. True total colour blindness (monochromacy) is very rare.

Most CVD is anomalous rather than total - the cone is shifted, not missing - so severity varies a lot from person to person, and many people with mild CVD do not know they have it. That has a design consequence: you cannot rely on users to 'declare' the problem or work around it. You also cannot rely on your own eyes to catch it, because if you have typical colour vision the failing pairs look perfectly distinct to you. This is why CVD design is done with rules and tools, not intuition.

One more honest point: CVD is not a disease to be 'fixed' and mostly is not correctable. The tinted glasses sold for it can boost some contrast for some people but do not restore normal colour vision. The designer's job is not to cure anyone - it is to build information that does not depend on the colour distinctions a large minority of people cannot make.

To design around CVD you need a feel for which colours collapse together, and the honest answer is not as simple as 'red looks green'. In red-green deficiency, the whole red-orange-yellow-green range gets compressed toward a band of muddy yellows and khakis, and separately, colours that look distinct to typical vision can converge. Some pairs that commonly confuse:

Commonly confused under red-green CVD:
  red        <-> green
  green      <-> brown / khaki
  red        <-> brown / dark grey
  blue       <-> purple  (purple loses its red)
  pink       <-> grey / pale blue
  teal       <-> grey
Under blue-yellow (tritan) CVD:
  blue       <-> green
  yellow     <-> pink / pale grey

Notice two traps. First, purple and blue merge for red-green viewers because purple is blue plus red and the red part fades - so a palette that distinguishes a 'blue zone' from a 'purple zone' by hue alone will fail. Second, value and saturation do not fail: a person with CVD sees lightness perfectly well. That is the escape route we will use throughout - separate your colours by value and by strength, not only by hue.

The most useful thing you can do is see it for yourself. Free CVD simulators (browser extensions, design-tool plug-ins, phone apps) re-render an image as a deuteran, protan or tritan viewer would experience it. Run your sign, chart, line drawing or app screen through a deuteranopia filter - the most common case - and the failures jump out: the two lines that were red and green become the same olive; the 'error in red' text vanishes into the body colour.

The two-minute CVD check:
  1. export the design as an image
  2. open a deuteranopia simulator
  3. can you still tell every meaningful thing apart?
  4. if not, colour was carrying the message alone

Simulators are approximations, not medical instruments, and they cannot reproduce the full range of anomalous vision - but they are more than good enough to catch the ordinary mistakes, and they turn an abstract worry into something you can see and fix in the same afternoon.

Everything in this lesson reduces to one durable rule worth writing on the studio wall: never let colour be the only carrier of information. WCAG states it as success criterion 1.4.1, 'Use of Color', and it applies just as well to a floor plan, a facade diagram, a control panel or a train map as it does to a web page. Colour is a wonderful accelerator of recognition - it lets the eye find the red line or the green exit fast - but it must always be backed up by at least one non-colour cue.

The backups are cheap and familiar once you look for them:

Redundant cues to pair with colour:
  text label ......... 'Exit', 'Line 2', '-8%'
  shape / icon ....... triangle vs square vs circle
  pattern / texture .. solid vs dashed vs dotted line
  position / order ... always the same slot, top-to-bottom
  value / tone ....... make colours differ in lightness too
  direct labelling ... name the line on the line itself

Concrete examples make it stick. A chart that distinguishes series by colour alone should also label each line directly or vary its dash pattern - then it survives greyscale printing, a washed-out projector, and every kind of CVD at once. A form that marks errors 'in red' should also add an icon and the word 'Error', so the message does not evaporate for the one-in-twelve. A services schematic that codes pipes by colour should also letter or hatch them. A wayfinding scheme that names zones by colour (Lesson 8.3) should give every zone a name and a number too.

There is a quiet bonus here that sells the discipline to sceptical clients: redundancy helps everyone, not only people with CVD. The labelled chart is clearer in a bad photocopy; the named metro line is easier to describe on the phone; the iconned button is faster to find under stress. This is the core insight of universal design - that accommodating the edge of the range usually improves the middle too. Designing for the one-in-twelve is not charity subtracted from good design; it is good design, and it costs almost nothing when you build it in from the start rather than retrofitting it after a complaint.

If colour still has to distinguish things - and it usually should, because it is fast and pleasant - then choose a colour-blind-safe palette: a set of colours that stay distinct across the common deficiencies. The trick, from the last two sections, is to separate them on more than hue.

Three principles build a safe palette. First, vary value as well as hue. If your categories differ in lightness, they survive both greyscale and CVD; a safe set often runs light-to-dark as much as around-the-wheel. Second, avoid the classic collision pairs for anything that must be told apart - especially red-versus-green and blue-versus-purple. If you need 'warning' and 'go', reach for orange or vermillion versus blue rather than red versus green. Third, keep the count low. The more categories you encode by colour, the harder every viewer's job; beyond about six or seven, colour alone stops working for anyone and you must lean on labels and position.

A widely used, freely available ready-made is the Okabe-Ito palette - eight colours (black, orange, sky blue, bluish green, yellow, blue, vermillion, reddish purple) specifically chosen to stay distinguishable for the common CVD types. It is a sound default for diagrams, maps and data graphics when you do not want to tune a palette by hand, and several design tools ship similar 'colour-blind-safe' swatch sets.

Building and checking a safe palette:
  - draft it, then view in greyscale (value gap?)
  - run it through deuteran + protan + tritan sims
  - test the actual pairs that must differ, at real size
  - add a non-colour cue anyway (label / shape / dash)
  - re-check after any 'let us make it prettier' edit

Two honest limits. First, no palette is safe for every possible vision - monochromacy defeats all colour coding, which is the final argument for the redundancy rule. Second, safe palettes constrain your aesthetics a little; that is a real trade-off, not a myth, and the answer is to spend the constraint where it matters (information, safety, wayfinding) and keep your free hand for decorative, non-load-bearing colour. The goal is not a drab world - it is a world where nobody is locked out of the meaning because of the colours they happen to see.