Three IQ Points: What the Evidence on Breastfeeding Benefits and Brain Development Actually Shows

The number is 3.44.

That is the IQ advantage usually attributed to breastfed children, and it is the single figure doing most of the work whenever breastfeeding benefits and brain development come up in a pediatrician's office or a comment thread. It comes from a meta-analysis by Horta, Loret de Mola, and Victora (2015) in Acta Paediatrica, which pooled observational studies of breastfeeding and later intelligence and found a difference of 3.44 IQ points, with a 95% confidence interval of 2.30 to 4.58. In the subset of studies that adjusted for the mother's own IQ, the estimate fell to 2.62.

That drop is the most interesting thing in the paper. It is almost never quoted.

What 3.44 points actually measures

IQ scales are built with a mean of 100 and a standard deviation of 15. So 3.44 points is roughly 0.23 standard deviations — in the neighborhood of the standard error of measurement for a single administration of a Wechsler test. Test the same child on two Tuesdays and you can see a swing that size for reasons having nothing to do with infant feeding.

It is also a shift in a group mean, not a property of any child. Two distributions overlapping almost entirely, with one nudged slightly right. Nothing in that number tells a specific family what would have happened to a specific baby.

And the groups are not randomly assembled. In high-income countries, breastfeeding duration tracks maternal education, household income, partner presence, and paid parental leave — every one of which independently moves childhood test scores. The statistical adjustments researchers apply are honest attempts to strip that out. They are also incomplete, and everyone who runs them knows it.

The sibling comparison that complicates it

Der, Batty, and Deary (2006), in the BMJ, worked with 5,475 children of mothers in the U.S. National Longitudinal Survey of Youth, including 3,161 sibling pairs. The raw gap looked like the familiar one: several IQ points in favor of breastfed children. After adjusting for maternal intelligence and home environment, it shrank to 0.52 points (95% CI −0.19 to 1.23) — statistically indistinguishable from nothing. Within sibling pairs, where the mother, the house, and the genome are largely held constant and only the feeding differs, the effect essentially disappeared.

Sibling designs have their own flaws. Mothers often feed a second child differently because of something about that child or that year — illness, work, a difficult first attempt. That is confounding too, pointing an unknown direction. But it is the cleanest natural experiment available, and it does not support a large cognitive effect.

The one randomized trial, and why it cuts both ways

The Promotion of Breastfeeding Intervention Trial (PROBIT) remains the strongest design in this literature. Kramer and colleagues cluster-randomized 31 Belarusian maternity hospitals and their affiliated clinics — 17,046 mother–infant pairs — to a Baby-Friendly-style breastfeeding promotion program or usual care. At 6.5 years (Kramer et al., 2008, Archives of General Psychiatry), the intervention group scored 7.5 points higher on verbal IQ and 5.9 points higher on full-scale IQ.

That is a big number, and it deserves two asterisks. The trial randomized promotion, not breastfeeding; and the pediatricians administering the tests were not blind to which clinics had received the intervention, with much of the effect concentrated among a subset of examiners. Teacher ratings of academic performance showed a far smaller gap. When the cohort was followed into adolescence, the cognitive advantage had largely faded.

So the best-designed study in the field points one way at age six and roughly nowhere at age sixteen.

Does breastfeeding make a baby smarter?

Probably by a small amount, and possibly not at all. The honest reading of the evidence is that breastfeeding is associated with something in the range of 2 to 3 IQ points, that a meaningful share of that is explained by which families breastfeed rather than by the milk, and that the residual effect — if real — is small enough to be invisible in any individual child. Anyone quoting a firm number is quoting a confidence the data does not have.

How milk could reach a neuron

If there is a real effect, it needs a physical route. The most-developed candidate runs like this, in the order it happens:

Docosahexaenoic acid (DHA) arrives in milk, typically 0.2–0.4% of total fatty acids and highly dependent on the mother's own diet. Infant lipases and bile salts release it in the small intestine. Enterocytes package it into chylomicrons; it enters lymph, then circulation. At the blood–brain barrier, a transporter called MFSD2A — characterized by Nguyen et al. (2014) in Nature — ferries DHA across in its lysophosphatidylcholine form. Inside, it is incorporated into the phospholipid membranes of synapses, where it alters membrane fluidity and receptor behavior. Oligodendrocytes, meanwhile, wrap axons in myelin, which is roughly 70–80% lipid by dry weight, and myelin is what lets electrical signals travel fast enough for the timing-dependent work of a developing cortex.

Deoni et al. (2013), in NeuroImage, imaged myelin water fraction in about 133 children and reported higher values in the exclusively breastfed. Small, observational, same confounds as everything above — but it is a real measurement of a real tissue property, which is more than most of this literature offers.

Here is the crack in the story. U.S. infant formula has contained added DHA since around 2002. A Cochrane review by Jasani and colleagues (2017) of long-chain polyunsaturated fatty acid supplementation in formula for term infants found no consistent neurodevelopmental benefit. If DHA were the mechanism, adding it should have closed the gap. It didn't clearly, which means either the mechanism is wrong, the delivery matters, or the gap was never mostly about milk.¹

What the number leaves out

It says nothing about cost. Borra, Iacovou, and Sevilla (2015) found, in the large British ALSPAC cohort, that mothers who intended to breastfeed and could not had elevated rates of postpartum depression — higher than mothers who never intended to. Sleep, feeding adequacy, and maternal mental health are outcomes too, and they are not on the IQ scale.

It also flattens the fact that the cognitive claim is the weakest well-known claim in this field:

An honest rule of thumb

Decide about feeding on the evidence that is solid — fewer infections now, and a genuinely large protective effect for a preterm baby — and treat the IQ number as a rounding error you are not obligated to optimize. If breastfeeding is working, the cognitive question shouldn't add pressure. If it isn't working, the cognitive question shouldn't be the reason you keep going at three in the morning.

One thing to try this week: the next time a headline promises a brain benefit from anything, open the study's abstract and find two facts — the sample size, and whether the analysis adjusted for maternal IQ or education. Both are usually in the second or third sentence. It takes ninety seconds and it will change what you believe about half of what you read.

The difference between three IQ points and zero IQ points is not something you will ever see in your own child, and any advice that depends on which one is true is advice built on sand.