What genetics can tell you about your baby's looks — and what it can't
The Punnett square you learned at fifteen has been quietly retired. What genetics can actually tell you about a future baby's face — a shorter list than anyone selling predictions would like.
The schoolbook model — brown beats blue, one gene — is wrong. Two blue-eyed parents can have a brown-eyed child.
A 195,000-person study found 124 variants across 61 genomic regions involved in eye color alone.
Genetics offers probabilities for single traits, never a portrait. The specific combination is decided randomly at conception.
Anything showing you a face is imagining it — sometimes honestly, sometimes not.
Mendel got fifty years of biology classes out of his pea plants, and the pea plants deserve it — round versus wrinkled really is one gene doing tidy arithmetic. Faces are not peas. The eye-color chart you filled in at fifteen has been quietly retired by the people who study this, and what replaced it says something useful about every app, chart, and grandmother currently predicting your future baby's looks.
The schoolbook chart is wrong
The classic story: brown dominates blue, one gene, draw the square. Geneticist John McDonald keeps a running debunk of this myth, and the observation that kills it is simple — two blue-eyed parents sometimes have brown-eyed children, which the single-gene model says is impossible. The real architecture surfaced in a genome-wide study of nearly 195,000 people published in Science Advances: 124 genetic variants across 61 regions of the genome, for eye color alone.
genetic variants involved in eye color — one trait. A face stacks thousands more on top.
What genetics can honestly offer
Probabilities, per trait, with wide error bars. Two brown-eyed parents will most likely have a brown-eyed child — likely, not certainly, because everyone carries variants they don't display, per MedlinePlus Genetics. Height, hair texture, dimples: same shape of answer, polygenic and probabilistic. What genetics cannot do, even with both parents' full genomes on a lab bench, is compose the portrait — which specific combination a child inherits is decided randomly at conception, one deal from a deck of billions.
Genetics offers probabilities for single traits, never a portrait. The deal happens at conception.
So who will the baby take after?
Nobody knows, and the one famous study claiming babies favor their fathers failed to replicate four years later — we wrote up both sets of receipts. The honest position: resemblance is real but unpredictable, and every tool showing you a future face is imagining, not predicting, whatever its marketing implies. The imagining can still be worth doing. A picture of a possible child asks you a concrete question — could I build a life with this person? — and asking it early is the entire reason Bean exists. The genetics stays unknowable. The feeling doesn't.
Fair questions
- Can genetics predict what my baby will look like?
- No. Genetics gives per-trait probabilities — likely eye color, likely range of height — but the specific combination a child inherits is decided randomly at conception. Even eye color alone involves 124 variants across 61 genomic regions.
- Can two blue-eyed parents have a brown-eyed baby?
- Yes. The single-gene eye color model taught in schools is wrong; eye color is polygenic, and combinations the old chart calls impossible happen regularly.
- Do AI baby generators use genetics?
- No. They compose a plausible child from the visible features in two photos using image-generation models. No DNA is read — and since genetics itself can't predict a face, the picture is imagined either way.
Elena Marsh — contributing editor. Elena dated on the apps through her twenties, met her partner at 31, and now writes the Bean Journal for people trying to make a similar choice with more care.
Date like it leads somewhere.
Bean shows you the family you could build before the first date. Three real choices a day.