The baby eye colour calculator above uses the two-gene Mendelian model that appears in most school biology courses. It is a genuine teaching model with a clear logic, and it is also an approximation that the underlying science moved past decades ago. Both of those things are true at once, and this page is built on saying so rather than hiding it behind a confident-looking percentage.
Arb Digital publishes free science calculators, and this one carries a stronger caveat than most. It cannot tell you what colour a particular child's eyes will be. What it can do is show how a simple dominance model distributes probability across three outcomes, and why a blue-eyed child from two brown-eyed parents is completely unremarkable rather than a mystery.
What This Calculator Does, and What It Cannot Do
It works two loci. The first carries a brown-determining allele that is treated as dominant over its alternative. The second carries a green-determining allele, and it only gets to express itself when the first locus has no brown allele present. A child with no brown allele and no green allele is modelled as blue. Everything else follows from ordinary Mendelian segregation.
Because a brown-eyed parent might be carrying a hidden non-brown allele, the tool has to make an assumption about how likely that is. It uses Hardy-Weinberg proportions from the two allele frequencies you can see and edit. That is why those fields are exposed rather than buried: change them and the answer moves, which is the clearest possible demonstration that the output is a model result rather than a fact about your family.
What it cannot do is predict an individual. It is not a genetic test, it carries no medical meaning of any kind, and it must not be used to draw conclusions about parentage or about any real person. If a question about inheritance in your own family matters to you, a clinical geneticist or a genetic counsellor is the right person to ask. Our Punnett square calculator takes the same position and is the better tool if you want to see the crosses laid out square by square for a single gene.
How to Use It
- Select each parent's eye colour from the three broad categories. Hazel is grouped with green, and grey with blue, because the two-gene model has no separate category for either.
- Leave the allele frequencies alone for a general answer. The defaults are illustrative values, not measurements of any specific population.
- Change the frequencies if you want to see how sensitive the result is. Raising the brown allele frequency makes brown-eyed parents more likely to be carrying two brown alleles, which reduces the chance of a non-brown child.
- Read all three probabilities, not just the largest. A 6 percent outcome is not an impossible outcome; it is one child in sixteen or so.
- Note the fourth figure, the chance the child's eyes differ from both parents. That is usually the number people are actually curious about.
How the Model Calculates the Probabilities
Each parent phenotype is compatible with several genotypes. A brown-eyed parent could be carrying two brown alleles or one brown and one non-brown, and could carry any combination at the green locus, since brown masks it entirely. The tool assigns probabilities to those possibilities using Hardy-Weinberg proportions, then computes the chance each parent passes each allele.
Once you have the transmission probability for each allele from each parent, the child's genotype probabilities are just products, and the phenotype follows the dominance rules. Blue requires two non-brown alleles at the first locus and two non-green at the second. Green requires two non-brown at the first and at least one green at the second. Brown requires at least one brown allele.
A worked example using the default frequencies of 0.35 and 0.55, with two brown-eyed parents. Conditional on being brown-eyed, each parent has about a 21 percent chance of carrying two brown alleles and 79 percent of carrying one, so each passes a brown allele about 61 percent of the time. The child therefore has two non-brown alleles with probability 0.394 squared, which is 15.5 percent. Of that 15.5 percent, roughly 80 percent inherits at least one green allele, giving 12.4 percent green and 3.1 percent blue, with 84.5 percent brown. Two brown-eyed parents having a blue-eyed child is an ordinary event in this model, not an anomaly.
Why the Two-Gene Model Is Wrong, and Still Useful
The single-gene version of this story — brown dominant, blue recessive, done — was the standard classroom explanation for most of the twentieth century, and it is simply incorrect. The National Library of Medicine's own genetics reference on whether eye color is determined by genetics states directly that researchers used to think eye colour followed a simple dominant-recessive pattern and that later studies showed the model was too simplistic. Two genes on chromosome 15, OCA2 and HERC2, do most of the heavy lifting, but at least eight further genes contribute measurably.
The mechanism is quantitative rather than categorical. Iris colour comes from how much melanin sits in the front layer of the iris and how light scatters through the tissue behind it. There is no blue pigment in a blue iris at all; the colour is a scattering effect, the same physics that makes the sky blue, revealed when there is little melanin to absorb the light first. Because melanin amount is continuous, so is eye colour, and any model with three boxes is forcing a spectrum into categories.
A two-gene model survives in teaching because it does one thing well: it explains, correctly, why recessive traits can skip generations and reappear. That single insight is worth having, and it is the reason this page exists. It just does not license a prediction.
What the Genetics Actually Says About Prediction
Modern forensic and research work does predict eye colour from DNA, and it does it far better than any parent-colour model can, because it reads the variants directly. A single variant in an intron of the HERC2 gene, rs12913832, regulates how much the neighbouring OCA2 gene is expressed and accounts for the largest share of the variation in populations of European ancestry.
Even that variant is not deterministic. A study published in the journal Genes in 2023 on the association between variants in the OCA2-HERC2 region and blue eye colour reported that 43 of 166 Norwegians carrying the genotype expected to produce brown eyes did not have brown eyes, and went looking for the additional variants that explained the discrepancy. If a direct genotype read gets it wrong a quarter of the time in one population, a model that starts from what two parents look like has no business being called a predictor.
Babies, and the Colour That Changes
One practical point the genetics does not cover. Many babies of European ancestry are born with eyes that look blue or grey and darken over the first year or two, because melanin production in the iris continues after birth and is triggered partly by light exposure. Babies of African and Asian ancestry are more often born with brown eyes already, since melanin deposition is further along at birth.
The colour typically settles somewhere between six months and three years, and it can carry on shifting subtly for longer. So an early observation is not the final answer, and a model's percentages are not falsified by a newborn's appearance. Both are describing the same underlying gradual process from different ends. Hazel, which sits awkwardly in every categorical model, is often the clearest case of this: it is a genuinely intermediate melanin level, not a distinct third category.
How to Read a Probability Without Over-Reading It
A percentage from any inheritance model describes a long-run frequency across many independent events, not a property of the next one. If the model says 12 percent green, the honest reading is that in a large group of couples with these characteristics, roughly one child in eight would have green eyes. It says nothing about which children.
Each child is also an independent draw. Three blue-eyed children in a row does not make the fourth more likely to be brown, in the same way that three coin flips do not influence the fourth. The probability calculator handles that kind of compound question directly, and the expected value calculator is useful when you want the average across several outcomes rather than a single one. If you want to see a single-gene cross drawn out properly, the Punnett square calculator shows the grid.
Arb Digital publishes hundreds of free calculators across biology, maths, physics and finance — no sign-up, no limits. If something you need is missing, tell us and we will look at building it.
Browse All Free Tools Suggest a ToolCommon Mistakes to Avoid
- Treating the largest percentage as the answer — a 62 percent outcome means 38 percent of children in that situation have something else, which is not a rare event.
- Assuming two blue-eyed parents cannot have a brown-eyed child — the simple model says they cannot, and the real genetics says it is uncommon but documented. That gap is the model's failure, not nature's.
- Reading a newborn's eye colour as final — iris melanin continues to be deposited after birth, and the colour often settles somewhere between six months and three years.
- Using any of this to reason about parentage — eye colour carries no evidential value whatsoever about who a child's parents are.
- Forgetting the frequencies are assumptions — allele frequencies differ hugely between populations, and the numbers this page defaults to are illustrative rather than measured.
Related Free Tools From Arb Digital
See a single-gene cross laid out properly with the Punnett square calculator, work through the ABO system with the blood type calculator, and handle compound chances with the probability calculator. The expected value calculator averages across outcomes, the Bayes theorem calculator updates a probability when new information arrives, and the DNA to mRNA converter covers transcription for coursework. The full free online tools hub lists everything else.
Frequently Asked Questions
Yes, and it is common rather than surprising. If both parents carry one brown allele and one non-brown allele, the simple model gives their child roughly a one in sixteen chance of blue eyes, and the real genetics allows it too because several genes are involved.
As a prediction for one child, not accurate at all. It applies a two-gene model to a trait shaped by at least ten genes, so it describes broad tendencies across many families and cannot be expected to be right about any individual.
No. The National Library of Medicine's genetics reference states that the older single-gene model was too simplistic. OCA2 and HERC2 on chromosome 15 have the largest effects, with at least eight other genes contributing measurably.
Because those numbers set how likely a brown-eyed parent is to be carrying a hidden non-brown allele. That is genuinely unknown from eye colour alone, so the model has to assume it, and the fields are editable to make that assumption visible.
Usually somewhere between six months and three years, because melanin continues to be deposited in the iris after birth. Many babies of European ancestry are born with eyes that look blue and darken later.
Awkwardly. Hazel is an intermediate amount of iris melanin rather than a distinct category, so a model with three boxes has to group it with green. That grouping is one of the model's clearer weaknesses.
No. Eye colour carries no evidential value about parentage at all, and this page must never be used for that purpose or for any medical, legal or family decision.
Because the colour comes from light scattering rather than pigment. With little melanin in the front layer of the iris, shorter wavelengths scatter back out, the same effect that makes the sky blue.
This calculator is an educational tool that applies a simplified two-gene Mendelian model to a trait that is genuinely polygenic. Every figure it produces is a modelled probability, never a prediction about a real child. It is not a genetic test, it carries no medical meaning, and it must not be used to draw any conclusion about parentage or about any individual person. Anyone with a question about inheritance in their own family should speak to a qualified clinical geneticist or genetic counsellor.
