The whole picture, brieflyPhysiology, prevalence, inheritance
One man in twelve. Almost none of them can explain it.Mechanism, prevalence and inheritance of colour vision deficiency
Colorblindness is common enough that you already know several people who have it, and odd enough that most of what gets said about it is wrong. So here is the real thing: what is broken, who gets it, why it runs in families, what the dot tests can and cannot tell you, and what can honestly be done about it.Colour vision deficiency affects approximately 8% of males of Northern European ancestry. This page covers the photoreceptor basis, prevalence by type and population, X-linked inheritance, the limits of pseudoisochromatic screening, comparative colour vision, and the current state of correction and gene therapy. Sources are named throughout.
Start with how manyPrevalence
Twelve marks, one of them shifted. Common enough to be in every classroom, rare enough that nothing is built with it in mind. An all-boys school of a thousand pupils in England has eighty-odd of them, and most have never been told.Twelve units at the male prevalence rate, one affected. Colour Blind Awareness gives the worked case of an English all-boys school of 1,000 pupils containing roughly 80 to 85 colour deficient students.
About one man in twelve has some form of red-green colorblindness, and about one woman in two hundred. Those figures are for people of Northern European descent, and they genuinely do move: in Chinese and Japanese men the rate is closer to one in twenty. Scandinavia runs higher than anywhere, at around one man in nine.Prevalence in European Caucasian populations is approximately 8% in males and 0.4 to 0.5% in females (Birch 2012, JOSA A). The same review reports 4 to 6.5% in men of Chinese and Japanese ethnicity, and rates near 4 to 5% in men of African ancestry. Colour Blind Awareness places Scandinavian male prevalence at 10 to 11%. Birch attributes the variation to founder events and genetic drift rather than to selection.
What is actually brokenPhotoreceptor basis
You have three kinds of colour sensor in each eye, tuned to long, medium and short wavelengths. Call them red, green and blue if you like. Your brain does not read any one of them directly; it works out colour by comparing what the three of them report. The red and green sensors are already tuned close together, which is the whole vulnerability, and in the common kind of colorblindness one of them drifts closer still. Nothing goes missing. The two reports just stop disagreeing by enough to be useful, and everything that depended on that disagreement gets vague.Normal trichromacy depends on three cone classes with peak spectral sensitivities near 420 nm (S), 534 nm (M) and 564 nm (L). Hue is derived from the ratios between their responses rather than from any single channel. In anomalous trichromacy a photopigment variant shifts one peak: in deuteranomaly the M peak moves toward the L peak, compressing the difference signal that carries red-green discrimination. All three classes remain functional, which is why chromatic perception persists in full while discrimination along one axis degrades.
- Blue sensorS cone, 420 nm
- Green sensorM cone, 534 nm
- Red sensorL cone, 564 nm
- Where red and green agreeM and L response overlap
The peaks are where they really are. The shape of each curve is drawn to a standard pigment template rather than traced off a spectrophotometer, because the thing worth looking at here is how much the red and green curves overlap before anything has even gone wrong.Peaks plotted at 420, 534 and 564 nm. Curve shape follows a visual pigment absorbance template (Govardovskii et al. 2000) rather than the measured Stockman and Sharpe fundamentals: the illustration is of relative overlap, not of absolute quantal sensitivity. The shifted curve is drawn at a 12 nm displacement, within the range reported for deuteranomalous observers.
Four kinds, at their real sizesDeficiency types and relative prevalence
Colorblind is not one condition, and the four red-green kinds are not evenly sized. Of those eight men in a hundred, five have the mildest one. That is why a site like this defaults to it, and why two colorblind people can disagree about what a colour looks like without either of them being wrong. Each row below runs the same six colours through that type at its real strength, using the same maths as everything else here.Of the approximately 8% of affected males: deuteranomaly ~5%, protanomaly ~1%, deuteranopia ~1%, protanopia ~1% (Colour Blind Awareness, consistent with Birch anomaloscope data). The classic ratio among affected men is 1 protanope to 1 protanomalous to 1 deuteranope to 5 deuteranomalous. Each row applies the Machado 2009 matrix for that type and severity to a common swatch set: the anomalous rows at severity 0.80, the dichromatic rows at 1.0.
The colours going inSource values, unmodified
Green-weakDeuteranomaly
Red-weakProtanomaly
No green sensorDeuteranopia
No red sensorProtanopia
One hillside, five times over. The swatches above prove it on paint; this is the same four kinds of eyes on a real afternoon. Watch the reds go first, then watch the two on the right lose the difference between a maple and a pine.A single source frame under the same four transforms applied to the swatch matrix above. Long-wavelength foliage converges toward the mid-spectrum first; in the dichromatic renderings the deciduous and coniferous canopy become separable by luminance and texture only.
- Green-weakDeuteranomaly
- Five men in a hundred. The green sensor has drifted toward red and the two crowd each other.Approximately 5% of males. M peak shifted toward longer wavelengths. The most common form of CVD.
- Red-weakProtanomaly
- One in a hundred. Same confusion, plus reds go dark: a red car in shade can read as almost black.Approximately 1% of males. Reduced L-cone sensitivity, peak shifted shorter. Adds a luminance effect absent in deuteranomaly: long-wavelength stimuli appear substantially darker.
- No green sensorDeuteranopia
- One in a hundred. Not weakened, absent. Red and green cannot be told apart at all.Approximately 1% of males. M cone class non-functional. Dichromacy: severity 1.0 in the Machado model.
- No red sensorProtanopia
- One in a hundred. Red is not just confusable now, it is dim. Red text on a dark background can vanish.Approximately 1% of males. L cone class non-functional. Long-wavelength stimuli lose most of their luminance contribution as well as their hue.
Two more kinds exist and both are rare. Blue-yellow colorblindness runs about one person in thirty thousand, and because it is not carried on the X chromosome it turns up in women as often as men. True total colorblindness, where there is genuinely no colour at all, is rarer still, and it brings problems the red-green kinds do not: poor detail vision, involuntary eye movement, and enough light sensitivity that dark glasses indoors are normal. If someone has been told they are totally colorblind by an optician with a book of dot plates, they were probably told wrong. That test cannot measure it.Tritan defects (OPN1SW on chromosome 7, autosomal dominant, therefore equal across sexes) and achromatopsia are each estimated near 1 in 30,000 to 50,000. Complete achromatopsia presents additionally with reduced visual acuity, nystagmus and photophobia, and is autosomal recessive; CNGA3 and CNGB3 account for up to 90% of cases (GeneReviews). Colour Blind Awareness notes that patients reported as totally colour blind on Ishihara screening frequently have severe red-green deficiency instead, since the series cannot assess achromatopsia at all.
What these pictures are notWhat the simulation does not claim
Every simulated picture on this site shows which colours stop being tellable apart. None of them is a photograph of what somebody else experiences, and none of them could be. Nobody has ever seen out of anyone else's eyes, and no amount of maths changes that.The transforms on this site model loss of chromatic discrimination. They identify which stimulus pairs become effectively indistinguishable for a given observer class. They are not renderings of subjective experience, which is not externally accessible and is not what the Machado model describes.
Overheard, in front of one of these pagesReported exchange
The reds have all gone tan on that side.Observer with normal trichromacy, viewing the transformed panel.
Has it? That still looks red to me.Deuteranomalous observer, viewing the same panel.
What about this, on its own. What colour is that?The same value presented in isolation, without surround or object context.
That is tan. Obviously.Deuteranomalous observer, same value, isolated presentation.
Neither answer was wrong. Two different things were going on at once. His eyes apply the same shift to both halves of the comparison, so the filtered picture looks to him a great deal like the unfiltered one, and he called it red because red is the word he has always used for that. Then the colour was lifted out and shown on its own, and everything he names colours by went with it: the object it belonged to, the green around it, the memory of what that thing is supposed to be. Left with a bare patch and nothing to go on, he read the patch.Two effects compound. First, an anomalous trichromat re-applies his own transform to both panels, so the difference across the divider is largely cancelled for him; the measured figures are below. Second, colour naming under compressed discrimination leans heavily on context: object identity, surround, and prior expectation. Removing the surround removes the inference, and the response reverts to a judgement of the isolated stimulus.
Measured on this site's own bright red, in the units colour scientists use for how different two colours look. Anything under about two is a difference you would struggle to see at all.CIE76 delta-E on the site's spec-red token at severity 0.80, computed at build time by src/lib/palette.ts. A delta-E near 2.3 is the conventional just-noticeable difference.
So this page will not tell you that a red tulip looks tan to a colorblind person. It does not, and anyone who says otherwise is guessing at something they cannot check. What the pictures show is which colours stop being separable, and that part is measurable, repeatable and worth knowing. The rest belongs to the person doing the seeing.Accordingly, no claim is made here about the phenomenal character of colour experience under CVD. The demonstrable content is the collapse of discrimination between specific stimulus pairs, which is measurable and reproducible; qualia are neither.
Why it is nearly always menX-linked recessive inheritance
The genes for the red and green pigments sit next to each other on the X chromosome. Men have one X, so a single faulty copy is the whole story. Women have two, and the working copy covers for the broken one, so a woman generally has to inherit it from both sides at once. That also explains the way it seems to skip a generation: a colorblind man gives his X to every daughter, so none of them are colorblind and all of them carry it, and then it surfaces again in their sons. A father cannot pass it to a son at all. Sons get his Y.OPN1LW and OPN1MW lie in a tandem array on the X chromosome and follow an X-linked recessive pattern (MedlinePlus Genetics). Hemizygous males express the phenotype from a single altered copy; females require alteration on both X chromosomes, and heterozygotes are typically unaffected carriers. Fathers cannot transmit X-linked traits to sons. All daughters of an affected male are obligate carriers, which produces the characteristic apparent skip: affected grandfather, carrier daughter, affected grandson. Blue-yellow deficiency is the exception, sitting on chromosome 7 and inherited autosomal dominantly.
| X | Y | |
|---|---|---|
| X | Daughter, in the clearDaughter, homozygous unaffected | Son, in the clearSon, unaffected |
| X | Daughter, carries itDaughter, heterozygous carrier | Son, colorblindSon, affected hemizygote |
One cross, four outcomes, and the useful thing is what is missing: no daughter of this pairing is colorblind, but half of them carry it onward. Half the sons have it outright.Carrier mother by unaffected father: 50% of sons affected, 50% of daughters carriers, no affected daughters. For a daughter to be affected she must inherit an altered X from both parents, requiring an affected or carrier mother and an affected father.
The arithmetic is neat enough to be worth doing yourself. If eight in every hundred X chromosomes carry it, then eight in every hundred men are colorblind, because one X is all they have. A woman needs two of them, so her odds are eight percent of eight percent: about one woman in a hundred and fifty. The measured figure is one in two hundred. Close enough that the model is clearly right.Given an allele frequency of approximately 0.08, male prevalence equals the allele frequency under hemizygosity, at 8%. Female prevalence is that value squared: 0.0064, or 0.64%, against an observed 0.4 to 0.5%. The agreement between the predicted and observed female rate is one of the cleaner confirmations of the inheritance model in human genetics.
How the circles of dots workPseudoisochromatic plate screening
Shinobu Ishihara published the dot plates in 1917 and they are still the standard screen. The number is painted in dots that differ from the background dots by hue and by almost nothing else: the brightness is deliberately jumbled across both sets, so you cannot cheat by looking for the lighter shape. If your hue channel is compressed along the axis that plate was built on, there is nothing left to read the number with. The dots do not hide anything. They just remove every cue except the one you are missing.Ishihara (1917) plates are pseudoisochromatic: figure and background dots are drawn from colours separated along a specific confusion axis, with luminance randomised across both sets to eliminate brightness as a segmentation cue. An observer whose discrimination is compressed along that axis cannot segment the figure. The clinical series runs to approximately 38 plates, presented in random order at roughly three seconds each, under illumination near 6000 to 7000 K, with tracing disallowed.
One of ours, generated rather than scanned. Every dot in it was checked against the same simulation code that runs the rest of this site.A generated plate from this site's own set. Figure and background colours are verified against src/lib/simulate.ts at build time. Not a reproduction of the Ishihara series.
The number in both circlesFigure 53
Two things it cannot do, both worth knowing before you trust a result. It screens for the red-green kinds only, so blue-yellow and total colorblindness walk straight past it. And even on its home ground it catches roughly nine in ten cases, not ten in ten. Other tests exist for the questions it cannot answer, which is why an actual optometrist does more than hand you a book.The Ishihara series screens red-green defects only and cannot assess tritan or achromatopsic vision; sensitivity for red-green defects is approximately 90 to 95%. The Hardy-Rand-Rittler plates and arrangement tests (Farnsworth-Munsell 100 Hue, Farnsworth D-15) address type and severity questions the Ishihara series is not designed for.
Take the colorblind test and see where you sitRun the pseudoisochromatic screening test
The chemist who left his eyes to scienceDalton, 1794: the first description
John Dalton, the atomic theory man, worked out in 1794 that he and his brother were seeing colour differently from everyone around them, and wrote the first scientific account of it. He was wrong about the cause. He believed the fluid inside his eye was tinted blue, and asked for it to be checked after he died. In 1844 they looked, and it was perfectly colourless. The answer had to wait until 1995, when someone sequenced DNA from his preserved eye tissue and found he had no green pigment at all. The condition was called Daltonism for a century and a half before anybody could say why he had it.John Dalton (1766 to 1844) presented the first scientific description of colour vision deficiency to the Manchester Literary and Philosophical Society in 1794, publishing "Extraordinary facts relating to the vision of colours" in 1798. He hypothesised a blue-tinted vitreous humour and requested post-mortem examination; the 1844 autopsy found the fluid colourless. DNA analysis of his preserved ocular tissue, published in Science in 1995, established him as a deuteranope lacking the M photopigment entirely. The term Daltonism first appears in print in 1827, via Prevost of Geneva.
- 1794 Writes the first description of itFirst scientific description, Manchester Lit and Phil
- 1844 The autopsy he asked for. His theory was wrong.Post-mortem examination: vitreous humour colourless
- 1995 DNA from the preserved tissue: no green pigmentMolecular analysis (Science, 1995): deuteranope
The island where it went the other wayPingelap: a founder effect
A typhoon hit Pingelap atoll in Micronesia around 1775 and left almost nobody alive. The survivors repopulated the island, and one of them carried the gene for total colorblindness, the genuinely rare kind where there is no colour at all. Today somewhere between four and ten islanders in every hundred have it and around thirty in a hundred carry it. Everywhere else on earth it is roughly one person in thirty thousand. Nothing about the gene changed. The population it was in got very small for one generation, and that was enough.A typhoon circa 1775 reduced the population of Pingelap atoll to a small number of survivors, at least one of whom carried an achromatopsia allele. Winick et al. (American Journal of Human Genetics, 1999) report 4 to 10% of the Pingelapese affected and approximately 30% carriers, against a background prevalence near 1 in 20,000 to 50,000. The locus was identified as CNGB3 on chromosome 8. The population is described for a general readership in Oliver Sacks, The Island of the Colorblind (1997).
That is the whole place, and that is the whole reason. A population that small, cut off by that much ocean, means one survivor with one gene can still be shaping how people see two centuries later.The atoll at scale. Population isolation of this degree is what allows a single founder allele to reach a frequency two to three orders of magnitude above its global background.
Almost nothing else sees what you seeComparative colour vision
Three sensors is not the standard model, it is just ours. Most mammals run two, which puts your dog roughly where a red-green colorblind human is, permanently. Birds run four, and the fourth reaches into ultraviolet, so there are markings on other birds that no human has ever seen. And then there is the mantis shrimp, which has up to sixteen and ought to win outright, except that when researchers actually tested it on similar colours it did worse than we do. More sensors turned out to buy it speed, not precision.Trichromacy is not the vertebrate norm. Most mammals, including dogs, cats and horses, are dichromats with red-green discrimination broadly comparable to a human deutan. Many birds, reptiles and fish are tetrachromats, frequently with a fourth class extending into the ultraviolet. Stomatopods carry up to 12 to 16 photoreceptor classes, yet behavioural discrimination testing (Thoen et al.) shows poorer performance on adjacent hues than trichromatic humans, indicating a fast coarse-encoding scheme rather than fine ratio comparison. Primate trichromacy is thought to have arisen approximately 30 million years ago, possibly under selection for detecting ripe fruit against foliage.
- YouHuman, trichromat 3
- Your dogDog, dichromat 2
- A pigeonBird, tetrachromat 4
- A mantis shrimpStomatopod 16
Sensor types, not sharpness. The shrimp is the proof that these are different things: sixteen kinds of receptor, and it still cannot separate two close colours as well as you can.Photoreceptor class counts. Discrimination performance does not scale with class count; the stomatopod case is the counterexample.
Roughly, and the roughly matters. A dog is a dichromat, but its two pigments do not sit exactly where a red-green colorblind human is missing one, so this is the nearest honest approximation rather than a picture of what your dog sees. What it does get right is the part everyone notices: the dog and the grass are the same colour, which is why the ball you threw is somewhere in that general direction.An approximation, not a simulation. Canine cone peaks lie near 429 and 555 nm, which is not the arrangement produced by removing a human L or M class, so the Machado dichromatic transform is the closest available rendering rather than a model of canid vision. The qualitative result, convergence of a warm coat with green vegetation, holds.
There is a human version of the fourth-sensor story, and it is stranger than the marketing suggests. Because these genes sit on the X chromosome and come in slightly different variants, roughly one woman in eight is walking around with the material for a fourth cone type. Having the fourth cone and actually seeing more colours turn out to be quite different things. A 2010 study tested a group of carriers and found exactly one woman who reliably did.Approximately 12% of women carry alleles for anomalous trichromacy and, through random X-inactivation, express four cone classes retinally (Jordan et al. 2010). Functional tetrachromacy is far rarer: that study identified a single subject, cDa29, who reliably demonstrated it under test. Retinal circuitry and the absence of relevant visual experience are the proposed limiting factors.
Can it be fixedCorrection and treatment
No, and be careful with anyone selling otherwise. Inherited colorblindness has no cure. It also does not get worse, which is the good half of the same fact: it is a wiring difference you were born with, not a disease progressing. What does exist is a set of tools, and the honest version of each is below.There is no cure for inherited colour vision deficiency. Because it is congenital rather than degenerative it does not progress (National Eye Institute). Management is adaptive. Two intervention classes are worth assessing on their evidence: spectral notch filters and gene therapy.
The glassesSpectral notch filters
They cut out the band of light where the red and green sensors overlap most, which pushes the remaining colours further apart. The evidence is genuinely split, and pretending otherwise would be doing the thing this site exists to correct. A 2018 study in Granada found they do not correct colorblindness and that wearers see the same colours differently rather than seeing new ones. Work published in 2020 and 2024 found real, measurable gains in colour contrast for mild and moderate cases. Both results can be true at once. They are an assistive tool: they help some people, they do nothing for the severe forms, and no filter has ever added a colour to anybody.Spectral notch filters attenuate the region of maximal L and M response overlap, increasing chromatic contrast between the remaining channels. The evidence is mixed. A 2018 University of Granada study (n=48) concluded the filters do not correct colour vision and redistribute discrimination rather than expanding it. Werner et al. (Current Biology, 2020) reported increased chromatic contrast responses persisting after two weeks of wear in mild to moderate anomalous trichromacy, and Somers and Bosten (Vision Research, 2024) found reduced discrimination thresholds while explicitly stating that no new colour experiences are created. Efficacy is condition and illumination dependent, absent in severe deficiency, and the FAA prohibits their use during vision certification.
The gene therapyGene therapy
In 2009 researchers gave two colorblind squirrel monkeys the missing pigment gene, and the animals gained red-green colour vision as adults. That was the surprise: nobody was sure a grown brain could take on a colour channel it had never had. Seventeen years later there has still never been a human trial for ordinary red-green colorblindness, and the reason is not really technical. This is not a blinding disease, and putting a virus under somebody's retina to improve their sock drawer is a hard risk to justify. Trials do exist for total colorblindness, which is a serious condition, and the published results so far are modest.Mancuso et al. (Nature, 2009) delivered a human L-opsin gene by adeno-associated virus to the retinae of two adult red-green colour blind squirrel monkeys, producing red-green discrimination and demonstrating adult cortical plasticity for a novel chromatic channel. No human trial for common red-green CVD has been registered as of 2026; the 2015 Avalanche Biotechnologies programmes (AVA-322 protan, AVA-323 deutan) were shelved. The limiting factors are risk-benefit, since subretinal delivery carries detachment and infection risk in a non-blinding condition, and the genetic complexity of the X-linked opsin array. Human trials do exist for achromatopsia (CNGA3, CNGB3): McKyton et al. (Current Biology, 2023) report positive but limited improvement, and AGTC halted its CNGA3 programme after observing no consistent clinical benefit.
Where all of this comes fromSources and caveats
The numbers here come from the National Eye Institute, the American Academy of Ophthalmology, MedlinePlus, Colour Blind Awareness, and the papers named in the text above. Two things worth flagging. Prevalence figures are for Northern European ancestry unless stated, because the rates really do differ by population and a single global average would hide that. And you will find lists of colorblind celebrities everywhere: those are almost always self-reported in an interview rather than clinically confirmed, so this page leaves them out.Principal sources: Birch, J. (2012), Worldwide prevalence of red-green colour deficiency, JOSA A 29(3):313-320; Mancuso et al. (2009), Nature; Jordan et al. (2010); Winick et al. (1999), AJHG; Kohl et al., GeneReviews; MedlinePlus Genetics; National Eye Institute; Colour Blind Awareness. Caveats: prevalence figures apply to European Caucasian populations except where stated; acquired deficiency prevalence (estimated near 3%) is poorly quantified; reported deficiency in public figures is generally self-reported rather than clinically verified and is omitted here.
Now go and find out where you sitScreening and demonstration
The rest of this site does the other half of the job. A short run of those circles will put you roughly into one of the groups above, and then it hands you a link that shows somebody else what it looks like from in here.The screening estimates deficiency type and severity band from responses to generated plates, and produces a shareable simulation configured to that result.
Find out what you seeTake the screening test Or just look aroundSkip to the simulator