When a flash fires straight into an eye, light passes through the tear film, cornea, lens and vitreous, bounces off the blood-rich retina at the back, and comes straight back out. That is the red-eye you spend your evenings editing out of family photos. Ophthalmologists call it the red reflex, and getting one at all means the whole optical path is clear.

Leukocoria is what you see when it is not. The AAO defines it as "the reflection of white light seen upon direct illumination of the fundus through the pupil, in contrast to the usual red glow." Anything between the front of the eye and the back of it — a cloudy lens, a mass, scarred retina, abnormal vessels — can send back white instead of red.

The important part for a parent is what that does not tell you. A white reflex says the light path is interrupted. It says nothing whatsoever about what is interrupting it.

The commonest cause of a white dot in a family snapshot is not disease at all. It is geometry. When a child is fixating roughly 15° off axis — glancing slightly toward their nose while the flash fires — the light lands on the optic nerve head instead of the retina. The optic nerve is pale. It reflects back white, from a completely healthy eye.

In one series of photographs of children with no eye disease, this physiologic leukocoria showed up in 0.48% of pictures. That sounds small until you consider how many photographs you take. Retinoblastoma, by contrast, occurs in roughly 1 in 18,000 births. Put those two numbers side by side and the arithmetic is reassuring: the large majority of white pupils that parents notice are artefacts.

Dr Evan Silverstein, a paediatric ophthalmologist who published two such cases in Pediatrics in 2025, put it plainly: "Leukocoria is not always retinoblastoma. It is not always a cataract. Sometimes it can pick up something very normal."

Parents who search this go straight to retinoblastoma. It is the frightening answer, and it is real, but it is not the most common one. Congenital cataract affects roughly 3 to 4 in every 10,000 live births in the US. Retinoblastoma occurs in about 1 in 18,000. Cataract is several times more common.

It is also time-critical in a way that surprises people. A visually significant congenital cataract generally needs surgery at 4 to 6 weeks of age if it affects one eye, and by 6 to 8 weeks if it affects both — because a baby's visual system is wiring itself, and an eye that cannot see during that window may never learn to. That deadline is measured in weeks, not months, and it applies whether or not anyone is worried about cancer.

The story everyone tells about leukocoria is a hopeful one: a parent spots a white dot on their phone, the cancer is caught early, the child is fine. The survival figures support the happy half of that. In the US, five-year survival for retinoblastoma is around 96%, and about 300 new cases are diagnosed each year, two-thirds of them before age two.

But survival of the child and survival of the eye are different questions. In a series of 1,265 patients, leukocoria was the presenting sign in 56.2% of cases — and it correlated with advanced disease. In a later series of 1,654 patients, children who presented with leukocoria had patient survival above 86% at five years but ocular survival of just 4% in unilateral cases. The lead investigator's conclusion was blunt: "Saving eyes and vision requires disease recognition before leukocoria."

That ocular figure is from before intra-arterial chemotherapy, which changed globe salvage dramatically — one centre now reports 96% ocular survival across 452 eyes. So the picture today is far better than those older numbers imply. But the underlying point stands: the white dot is a late signal, not an early-warning system. Treating your camera as a screening tool is a misunderstanding of what it is doing.

The American Academy of Pediatrics recommends a red reflex examination before discharge from the newborn nursery and at every subsequent well-child visit. Most parents assume this is a reliable net. It is not.

A meta-analysis of five studies covering 8,713 infants compared the undilated red reflex screen against a full dilated ophthalmological exam. Sensitivity for any ocular pathology was 7.5%. For pathology serious enough to require intervention, 17.5%. Among the conditions the screen missed was a retinoblastoma. The authors' own conclusion: "a normal red reflex finding during screening does not exclude ocular disease."

There is a second gap. The red reflex lives in a 2008 AAP policy statement, but it is not named in the vision footnote of the Bright Futures periodicity schedule that structures most well-child visits. It is recommended; it is not always scheduled. We sell eyewear, and we are telling you that a clean bill of health from a well-child visit is weaker evidence than it feels.

There is a widely repeated protocol for photographing your child's eyes: dim the room, force the flash, switch red-eye reduction off. The first two are sound — analysis of 52,982 photographs found a flash had been used in all but two of the leukocoric images. The third is not supported, and the best available data points the other way. In one dataset roughly 25% of leukocoria-containing photographs were taken with red-eye reduction active; in another, the feature "did not entirely inhibit leukocoria." It is plausible physics that nobody has demonstrated.

You will also see claims that HDR and computational photography hide the reflex. We looked for a source for that and found none — not from any ophthalmological, paediatric or imaging authority. We are leaving it out rather than repeating it.

Two more things worth knowing before you trust an app. Most of the "photos catch it early" evidence — including the striking finding that leukocoria was visible a mean of 484 days before diagnosis — comes from research published by the inventors of a commercial screening app. That does not make it wrong, and the finding is genuinely important, but you should know who produced it. And when an independent group tested that app against 34 eyes, it flagged only four, all late-stage. In children under two its specificity was 20% — meaning four out of five healthy children were flagged.

1. Look back through your photos. You are not looking for one white dot. You are looking for a pattern: the same eye, more than once, ideally when the child was looking at the camera. Keep the pictures — they are useful to the doctor.

2. Call, and say the word. "I have seen leukocoria in photographs of my child" gets a different response from "his eye looks a bit funny in pictures." US bodies are unanimous on urgency: the AAP says refer immediately, AAPOS and the AAO both say right away. In the UK, NICE is more measured, advising clinicians to consider urgent referral for assessment within two weeks where the red reflex is absent. Both are defensible; the American guidance is the more cautious.

3. Ask for a dilated exam, not a red reflex check. This is the single most useful sentence in this article. Given the numbers above, a repeat red reflex test is not the reassurance you came for. What settles it is a paediatric ophthalmologist looking at the back of the eye with the pupil dilated.

4. Expect it to be nothing. Most of the time it will be, and that is not a wasted appointment. It is the only way anyone can tell.

If an eye that turns is what you have actually noticed, read strabismus in kids — it is the second most common presenting sign and far more common overall. Lazy eye is another condition children never report themselves, and what a school vision screening misses covers the same gap between screening and a real exam that this article is about. If you are booking a first appointment, here is what actually happens at one, and 7 signs your child needs glasses covers the everyday things worth watching for.

AAO EyeWiki: Leukocoria · AAPOS: Leukocoria · AAPOS: Retinoblastoma · NCI PDQ: Retinoblastoma Treatment · National Eye Institute: Retinoblastoma · AAP: Red Reflex Examination in Neonates, Infants and Children · JAMA Ophthalmology: diagnostic accuracy of the red reflex test · AAO EyeWiki: Cataracts in Children · AAO EyeWiki: Coats Disease · Canadian Family Physician: pseudoleukocoria from off-axis fixation · PLOS One: leukocoria in longitudinal photographs · Eye: independent evaluation of smartphone leukocoria screening · NICE NG12: Suspected cancer recognition and referral