# How Baby Eye Color Is Inherited: OCA2, HERC2 and Why Two Brown-Eyed Parents Can Have a Blue-Eyed Baby _The two-gene model, the Punnett square that goes with it, and the 61 genomic regions that show why the square is only half the story._ - Source: https://proba.baby/en/blog/how-baby-eye-color-is-inherited - Published: 2026-09-05 - Author: ProbaBaby Team - Category: Genetics & Science - Tags: inheritance, eye-color - Language: English > The one-gene eye colour chart from school biology is wrong. Two genes on chromosome 15 do most of the work, dozens more fine-tune it, and that is exactly why two brown-eyed parents can have a blue-eyed baby. > **Medical disclaimer.** This article is for general information and does not replace medical advice or genetic counselling. Eye-colour prediction is entertainment, not a diagnostic test. If an inherited condition runs in your family, or if your baby's eyes look very different from each other, speak to your doctor. Baby eye color genetics is not the single-gene chart from school biology. Two brown-eyed parents can absolutely have a blue-eyed baby — roughly one child in four when both parents carry a hidden blue-associated variant. Two blue-eyed parents can have a brown-eyed child too, though that one is uncommon. The reason is that eye colour is mostly controlled by two neighbouring genes on chromosome 15, **OCA2** and **HERC2**, with dozens of other genes adjusting the final shade. Here is how that machinery works, and where a Punnett square stops being reliable. ## Eye colour is melanin, not pigment mixing There is no blue pigment in a blue iris. There is one pigment — melanin — and the only question is how much of it there is. The NEST researchers put it as plainly as it can be put: iris colour "appears to vary by the quantity of melanin within the iris, with brown irides simply possessing more melanin than blue irides." A lot of melanin absorbs most of the light that enters, and the eye reads as brown. Very little melanin, and the light scattered back out of the iris reads as blue. In between sits everything else: grey, green, hazel and amber are intermediate amounts of melanin. That single fact explains a lot of the confusion. Because it is one dial rather than a colour mixer, "which colour wins" is the wrong question. The right question is: **how much melanin will this baby's iris end up making?** ## What OCA2 and HERC2 actually do **OCA2** carries the instructions for a protein called the P protein, which sits in the melanocytes that make pigment. MedlinePlus is careful about what is known here: the exact function of the P protein is still unknown, but it is "essential for normal pigmentation and is likely involved in the production of melanin," and it "may transport molecules into and out of structures called melanosomes (where melanin is produced)." The chain from there is stated outright, though: common variations in OCA2 "reduce the amount of functional P protein that is produced. Less P protein means that less melanin is present in the iris, leading to blue eyes instead of brown." **HERC2** is the neighbour, and it does something less obvious. Buried inside it, in a stretch of DNA called intron 86, is a small regulatory switch that controls how strongly OCA2 is expressed — effectively an on/off dial for the gene next door. In 2008, Sturm and colleagues pinned the switch down to a single letter of DNA, a variant known as rs12913832. The authors concluded that the conserved region around it "represents a regulatory region controlling constitutive expression of OCA2," and that the blue-associated allele "leads to decreased expression of OCA2, particularly within iris melanocytes." Less OCA2, less melanin, blue eyes. That one SNP is remarkably powerful. In a sample of roughly 3,000 Australian adolescent twins and their siblings from southeast Queensland, more than 95% of northern European origin, it "predicted eye color significantly better (ordinal logistic regression R2 = 0.68, association LOD = 444)" than the best marker anyone had found before. That R² is a model-fit score rather than a clean "68% of eye colour explained," but the ranking it produces is unambiguous: almost every "eye colour gene" article you have read is really about this one switch. ## The Punnett square, done honestly Because that switch behaves close to dominant/recessive, you can draw the classic square for it. Call the brown-associated version **B** and the blue-associated version **b**. Two brown-eyed parents who each carry one of each — genotype **Bb** — produce this: | | **B** | **b** | |---|---|---| | **B** | BB — brown | Bb — brown | | **b** | Bb — brown | bb — blue | Three of the four combinations give brown eyes; one gives blue. So two brown-eyed parents who both carry the hidden variant have about a **25% chance** of a blue-eyed child, and each of their brown-eyed children has a two-in-three chance of being a carrier in turn. Nothing skipped a generation and nothing went wrong — the variant was simply invisible in the parents. Run the other crosses and you get the rest of the folk wisdom, corrected: - **Blue × blue (bb × bb)** — usually blue, because neither parent has a B to pass on. But MedlinePlus is explicit that brown-eyed children of two blue-eyed parents do occur — more than one gene feeds iris melanin, so this switch does not get the last word. - **Brown (BB) × blue (bb)** — every child is Bb: a carrier, and expected to be brown-eyed. This is the generation where blue "disappears" before resurfacing later. - **Brown (Bb) × blue (bb)** — a straight 50/50. ## Why the square is only half the story Now the honest caveat, and it is a big one. Blue-versus-brown is roughly binary; the actual range of human eye colour is a continuum, and the continuum is polygenic. In 2021, Simcoe and colleagues ran a genome-wide association study on 192,986 European participants plus 1,636 Asian participants. They found **124 independent associations arising from 61 discrete genomic regions**, including 50 loci that had never been reported before. Some of those genes affect pigmentation; others affect the structure of the iris itself. Together, common variants explained **53.2%** of eye-colour variation (95% confidence interval 45.4–61.0%) on a three-category colour scale. The paper notes that earlier studies had accounted for between 26% and 50%, so this is the most complete model so far — and still only about half. Their conclusion was blunt: the genetic complexity of human eye colour "considerably exceeds previous knowledge and expectations." Beyond OCA2 and HERC2, MedlinePlus names ASIP, IRF4, SLC24A4, SLC24A5, SLC45A2, TPCN2, TYR and TYRP1 as contributors. That is why green, hazel and amber refuse to fit a two-by-two grid: they are what you get when several dials land in the middle at once. A two-gene Punnett square gives you a sensible probability for *brown versus blue*. It cannot tell you whether the light eye that results will read as grey, green or blue-hazel in daylight. ## Newborn eyes: what the colour on day one really means Less than the internet implies. In the NEST study's own words, "changes in melanin production over time can yield changes in iris color from infancy" — so a newborn's colour is a starting point, and for some children it is not the final answer. The Newborn Eye Screening Test study imaged 192 newborns and found brown eyes in **63.0%**, blue in **20.8%**, green or hazel in **5.7%**, and indeterminate in **9.9%**. The numbers tracked closely with ancestry: among white infants, blue was the most common colour at 54.7%, while 80.0% of Asian infants and 77.8% of Native Hawaiian and Pacific Islander infants had brown eyes at birth. The "all babies are born with blue eyes" line does not survive contact with the data: it was never true globally, and even among the white infants whose experience produced the saying, blue was about half the group rather than all of it. Change afterwards is real, and it has been measured. The NEST authors cite the Louisville Twin Study, in which **10% to 20% of children changed iris colour between three months and six years of age**, and a subgroup of 10% to 15% of Caucasian subjects kept changing into adulthood. Read the same number the other way round and it is just as useful: at least four children in five kept the colour they had at three months. What none of these studies gives you is a birthday at which the answer is locked in — so treat "it settles by their first birthday" as folklore rather than a finding. ## Where grandparents come in Grandparents do not contribute DNA to your baby directly — everything arrives through the parents. What they contribute is **evidence about what the parents are carrying.** A brown-eyed father whose own mother had blue eyes must be Bb under this model; she could only have passed him a b. That one fact does not by itself make a blue-eyed baby 25% likely — that number needs *both* parents to be carriers — but it removes half the uncertainty. If the mother is a carrier too, you are in the square above; if she carries two brown-associated copies, blue is off the table for this model. The same logic works in reverse: two brown-eyed parents with four brown-eyed parents of their own, and no light eyes anywhere in the family, are more likely to carry two brown-associated copies each, and the blue outcome gets less likely. Family history is not a different mechanism — it is a way of narrowing down which square you should be drawing in the first place. ## How ProbaBaby helps ProbaBaby's [baby trait prediction](/ai-prediction) uses the same model this article describes for eye colour: a **two-gene calculation (HERC2 and OCA2), each resolved with a Punnett square**, producing probabilities across six modelled phenotypes — brown, amber, hazel, green, grey and blue. Those probabilities are computed in code, not written by a language model; the AI is handed the finished numbers and only explains them in plain language. The engine also models age progression for eye colour, because babies often start blue-grey, and it adjusts allele frequencies when you supply ancestry information. The Premium mode is the one that collects grandparent eye and hair colour, which is why it can reach a higher confidence score — confidence starts at 60%, rises with the family data you provide, and is hard-capped at 95%, since genetics never gives certainty. The exported PDF report shows a Punnett square for each candidate cross, so you can check the working rather than take the answer on trust, and it carries the app's own line: an entertainment estimate, not genetic testing or medical advice. You can read more about [the science behind the predictions](/genetics-science), or follow the pigment story alongside your [week-by-week pregnancy timeline](/pregnancy/week-by-week). ## FAQ **Can two brown-eyed parents have a blue-eyed baby?** Yes. If both parents carry one brown-associated and one blue-associated variant at the HERC2/OCA2 region, roughly one child in four is expected to inherit two blue-associated copies — the genotype most strongly linked to blue eyes. The variant was simply invisible in both parents. **Can two blue-eyed parents have a brown-eyed child?** It is uncommon, but MedlinePlus Genetics states it plainly: parents with blue eyes can have children with brown eyes. Because more than one gene contributes to eye colour, the main blue-brown switch does not have the final say on its own. **Which genes control eye colour?** OCA2 and HERC2, neighbouring genes on chromosome 15, account for most blue-versus-brown variation. MedlinePlus also lists ASIP, IRF4, SLC24A4, SLC24A5, SLC45A2, TPCN2, TYR and TYRP1 as contributors to the full spectrum. **How much of eye colour does the HERC2 variant explain?** Enough to dominate the trait. In the 2008 study that pinned it down, rs12913832 predicted eye colour with an ordinal logistic regression R2 of 0.68 in about 3,000 Australian adolescent twins and their siblings, more than 95% of them of northern European origin — a model-fit score, not literally 68% of the variation. Across the full colour range, a 2021 study of almost 195,000 people found common variants explain 53.2%. **Will my newborn's blue eyes stay blue?** Often they do, but not always. The NEST paper cites the Louisville Twin Study, in which 10% to 20% of children changed iris colour between three months and six years of age, and a subgroup of 10% to 15% of Caucasian subjects kept changing into adulthood. No study cited here names an age at which eye colour is finally fixed. **Do grandparents affect a baby's eye colour?** Not directly, but they are useful evidence. Under the two-gene model, a blue-eyed grandparent means a brown-eyed parent must carry a hidden blue-associated variant, which changes the odds you would calculate for the baby. **Is a green-eyed baby possible if neither parent has green eyes?** It is possible. Green and hazel are intermediate amounts of iris melanin, and MedlinePlus Genetics notes that several genes combine with OCA2 and HERC2 to produce a continuum of eye colours. A two-gene brown-versus-blue square cannot rule green out. ## Sources - MedlinePlus Genetics (NIH): Is eye color determined by genetics? — Updated 7 July 2022; OCA2, HERC2 intron 86, and eight further contributing genes: https://medlineplus.gov/genetics/understanding/traits/eyecolor/ - Sturm RA, Duffy DL, Zhao ZZ, et al. A single SNP in an evolutionary conserved region within intron 86 of the HERC2 gene determines human blue-brown eye color. Am J Hum Genet. 2008;82(2):424-431 — rs12913832; ordinal logistic regression R2 = 0.68 in ~3,000 Australians of northern European ancestry; regulatory control of OCA2 expression: https://pmc.ncbi.nlm.nih.gov/articles/PMC2427173/ - Simcoe M, Valdes A, Liu F, et al. Genome-wide association study in almost 195,000 individuals identifies 50 previously unidentified genetic loci for eye color. Sci Adv. 2021;7(11):eabd1239 — 124 independent associations across 61 genomic regions; common SNPs explain 53.2% of variation: https://pmc.ncbi.nlm.nih.gov/articles/PMC7946369/ - Ludwig CA, Callaway NF, Fredrick DR, Blumenkranz MS, Moshfeghi DM. What colour are newborns' eyes? Prevalence of iris colour in the Newborn Eye Screening Test (NEST) study. Acta Ophthalmol. 2016 — Acta Ophthalmol 94(5):485-488; 192 newborns: 63.0% brown, 20.8% blue, 5.7% green/hazel, 9.9% indeterminate: https://pmc.ncbi.nlm.nih.gov/articles/PMC4956505/ - MedlinePlus Genetics (NIH): OCA2 gene — Updated 13 May 2022; the P protein, melanosomes, and melanin production: https://medlineplus.gov/genetics/gene/oca2/ --- ProbaBaby (listed on the App Store as “Pregnancy & Baby Tracker - PB”) is a free iPhone, iPad and Apple Watch app for pregnancy, newborn and postpartum tracking, built for twins, in 32 in-app languages, with an optional AI baby-face prediction that computes Mendelian genetics (Punnett squares) from the traits you enter before any image is generated. Tracking, tools, PDF reports and family sharing are free; only AI predictions use credits — 5 starter credits cover your first Scientific prediction — and no subscription is required. Requires iOS 26 or later. Not available on Android.