Does Breastfeeding Make Children Smarter? The Evidence Behind the IQ Claim

The question

Few claims in parenting are stated as confidently as this one: breastfed babies grow into smarter children. The claim is not invented. Large observational studies consistently find that children who were breastfed score higher on intelligence tests, often by 3 to 7 points. A 2008 randomized trial in Belarus was widely reported as strong evidence of a causal effect. A 2015 meta-analysis put the average advantage at 3.44 IQ points, and a 2025 analysis of 373 outcomes in a British cohort reported full-scale IQ differences of about 4 to 5 points at ages 8 and 15.

But the same literature contains a puzzle. The randomized trial's full-scale IQ estimate had a confidence interval that ran from below zero to more than twelve points. Its 16-year follow-up found no significant intention-to-treat effect. Sibling comparisons, which hold constant the family environment that produces most of the confounding, find effects near zero. The newest genetic analysis finds no causal effect on intelligence. If breastfeeding really raised IQ by half a standard deviation, these designs should not disagree so sharply.

This brief asks the narrow question: how much of the observed association between breastfeeding and intelligence is causal, and what kind of effect is it? The short answer is that the association is real, the causal effect is probably small — on the order of zero to three full-scale IQ points, with a somewhat larger verbal component — and most of the raw gap reflects who breastfeeds rather than what breast milk does.

Why it matters

Feeding decisions are made under pressure. New parents are told that breastfeeding is the best start, that it protects against infection, and that it may make their child smarter. The first two claims have a solid evidence base; the third is the one that carries the most emotional weight. Patnode et al. (2025) reviewed the health literature and found associations with reduced respiratory and gastrointestinal infections, otitis media, asthma, obesity, and several other outcomes. If the cognitive benefit is small, the case for breastfeeding should rest on those health benefits, on support for mothers, and on the fact that feeding is only one part of responsive care. Overstating the cognitive benefit can make a difficult decision feel like a verdict on a parent, and it can push women who cannot or do not breastfeed toward guilt that the evidence does not justify.

The claim also matters for science communication. A single randomized trial with a wide confidence interval became "breastfeeding makes children smarter," and the correction has been slow. The result is a public conversation in which the strength of the claim and the strength of the evidence have drifted apart. The useful question is not whether breastfeeding is good — it clearly has short-term health benefits — but whether the specific promise of a cognitive boost survives the designs that can separate milk from mothering.

There is a policy dimension as well. Child development is shaped by many inputs: nutrition, infection control, responsive caregiving, education, income, and stress. If breastfeeding's cognitive effect is small, then presenting it as a major lever can distract from interventions with larger and better-established effects. That does not make breastfeeding unimportant. It makes the claim more precise.

What the standard claim says

The observational literature is large and consistent. A 2015 systematic review and meta-analysis of 17 studies found that breastfed subjects scored 3.44 IQ points higher (95% CI 2.30 to 4.58). The subset of studies that controlled for maternal IQ found a smaller advantage, 2.62 points (1.25 to 3.98). A 2023 review of 23 studies reached a similar conclusion, reporting a modest dose-dependent increase of 0.19 to 0.96 points per month of breastfeeding. The Pelotas 1982 birth cohort in Brazil found that participants breastfed for 12 months or more scored 3.76 points higher at age 30 than those breastfed for less than one month. The UK Millennium Cohort Study found a 0.26 standard deviation verbal advantage at age 14 for breastfeeding for 12 months or more versus never. A 2025 ALSPAC analysis reported full-scale IQ differences of 4.11 points at age 8 and 5.12 points at age 15.

These are associations, not causal estimates. The strongest observational studies adjust for maternal education, family income, birth order, and sometimes maternal cognitive ability. Adjustment typically cuts the estimate by half or more. The raw difference is not the causal effect, and the adjusted difference is still not the causal effect if the adjustment is incomplete.

The studies also measure different things. Pelotas used the Wechsler Adult Intelligence Scale at age 30. The Millennium Cohort used standardized verbal and spatial scores at five ages. The US National Longitudinal Survey of Youth used the Peabody Individual Achievement Test, an achievement measure rather than a full cognitive battery. ALSPAC used a broad battery with many outcomes. A 3-point difference on one instrument is not the same as a 3-point difference on another, and a global IQ score can hide domain-specific effects.

The randomized test

The strongest design is the Promotion of Breastfeeding Intervention Trial (PROBIT). It randomized 31 maternity hospitals in Belarus, not individual mothers, to a breastfeeding promotion program modeled on the Baby-Friendly Hospital Initiative. The trial enrolled 17,046 healthy infants; 13,889 were followed at age 6.5. The intervention raised exclusive breastfeeding at three months from 6.4% to 43.3% and increased any breastfeeding through 12 months. The comparison, therefore, was not breast milk versus formula. It was a hospital culture that promoted longer and more exclusive breastfeeding versus usual care in a setting where breastfeeding was already common.

At 6.5 years, the cluster-adjusted differences were +7.5 verbal IQ (95% CI 0.8 to 14.3), +2.9 performance IQ (-3.3 to 9.1), and +5.9 full-scale IQ (-1.0 to 12.8). The authors concluded that the trial provided strong evidence that prolonged and exclusive breastfeeding improves cognitive development. The full-scale interval is the problem: it included zero. The verbal interval excluded zero, but the clusters were large and the effective sample size was much smaller than the 13,889 children followed.

At 16 years, 13,557 participants were followed. The prespecified intention-to-treat analysis found no statistically significant differences in seven cognitive domains or a global score. After additional baseline adjustment, verbal function was +1.4 (0.3 to 2.5) and memory was +1.2 (0.01 to 2.4). The instrumental-variable analysis, which tries to estimate the effect of actually being exclusively breastfed for at least three months, found +3.5 verbal (0.9 to 6.1) and no differences in the other domains.

Two lessons follow. First, the effect shrinks with age and is domain-specific: verbal and memory, not performance or general ability. Second, the intention-to-treat estimate is the effect of the promotion program, not of breast milk. The instrumental-variable estimate is closer to the biological question, but it depends on the assumption that the program affected cognitive outcomes only through breastfeeding. That assumption is plausible but not proven, because promotion may also have changed maternal behavior and health-care use.

The trial also has an external-validity limit. Belarus had high breastfeeding initiation and a particular health-care system; the control group was not formula-fed from birth. The intervention changed the intensity and duration of breastfeeding, not the presence or absence of breast milk. A trial that randomized infant feeding itself would be ethically and practically impossible. That is why the observational and sibling literatures still matter.

The sibling test

Within-family comparisons ask a different question: when two children born to the same mother were fed differently, did the breastfed child score higher? This design removes stable family confounders — maternal cognitive ability, education, income, parenting style, neighborhood — that inflate between-family estimates. It does not remove confounders that differ between siblings, such as birth order, spacing, pregnancy complications, or differential maternal investment.

Der, Batty, and Deary (2006) used the US National Longitudinal Survey of Youth: 5,475 children of 3,161 mothers. Maternal cognitive ability was a stronger predictor of breastfeeding than race, education, age, poverty, smoking, home environment, birth weight, or birth order. Before adjustment, breastfeeding was associated with about 4 IQ points. After full adjustment, the difference was 0.52 points (95% CI -0.19 to 1.23), not significant. Among 332 sibling pairs discordant for breastfeeding status and 545 discordant for duration, none of the differences was significantly different from zero.

Colen and Ramey (2014) repeated the exercise with 25 years of NLSY data and 11 outcomes. In standard regression, breastfed children did better on 10 of 11. Within families, the estimates fell sharply and lost significance for all but one outcome. Evenhouse and Reilly (2005) used 16,903 adolescents from the Add Health study, including 2,734 sibling pairs. Most associations disappeared within families; the exception was a small cognitive association — 0.16 percentile points on the Peabody Picture Vocabulary Test per month of breastfeeding, significant only at the 10% level.

The sibling literature is not unanimous. Evenhouse and Reilly's positive result is real but weak; Der et al. and Colen and Ramey find nothing. The combined message is that the sibling designs rule out a large general-ability effect more than they prove an effect of zero. A 1- or 2-point effect could easily hide in samples of this size, especially with a noisy achievement test as the outcome. Sibling designs also depend on there being enough discordant pairs, and the children were not randomly assigned to feeding.

The genetic test

Two genetic strategies have been used. The first tests a gene-by-environment interaction. Caspi et al. (2007) reported that breastfeeding was associated with a 5.6- to 6.3-point IQ advantage, but the advantage was concentrated among children carrying a particular FADS2 variant, a gene involved in converting precursor fatty acids into long-chain polyunsaturated fatty acids. The idea was biologically plausible: breast milk supplies DHA and ARA, and children with a less efficient endogenous pathway might benefit more.

A collaborative meta-analysis with more than 12,000 individuals (Hartwig et al., 2019) found no strong evidence of that interaction. The pooled interaction coefficients were 0.12 (95% CI -0.19 to 0.43) and 0.06 (-0.16 to 0.27) for two FADS2 variants. A secondary subgroup analysis hinted that longer breastfeeding might matter, but the confidence intervals were wide and the finding was not robust. The replication failure is a useful caution: a plausible mechanism and a positive first study are not enough.

The second strategy is Mendelian randomization. Lu et al. (2025) used 49 genetic variants associated with breastfeeding as instruments and found no significant causal effect on intelligence (P = 0.165), nor on several other mental-health outcomes. The same study reported a small protective association with asthma. The intelligence null is consistent with the sibling and 16-year trial results, but it is not decisive. Breastfeeding is a behavior, not a molecule; genetic variants that predict it may also affect maternal education, mental health, or body weight, and those traits can influence child cognition. The instruments may also be weak. A null from an invalid instrument is not proof of no effect.

Taken together, the genetic evidence weakens the strong causal claim. It does not replace the randomized trial, but it adds to the pattern: when the confounding is reduced, the effect gets smaller.

The observational evidence that survives

Two large observational studies still suggest a modest effect. The Pelotas 1982 birth cohort followed 3,493 participants to age 30. Compared with breastfeeding for less than one month, breastfeeding for 12 months or more was associated with +3.76 IQ points (2.20 to 5.33), +0.91 years of education, and +341 Brazilian reals per month in income. The setting was chosen because breastfeeding was not strongly patterned by socioeconomic position, which reduces one major source of confounding. The association was dose-dependent, and a mediation analysis attributed 72% of the income effect to IQ.

The authors were explicit about the limits: they did not measure childhood home environment, maternal-infant bonding, or maternal cognitive ability. Those omissions matter. A mother who breastfeeds for a year may also read more, talk more, and provide more stimulation; the analysis cannot separate milk from mothering. The income and education findings are even more vulnerable to residual confounding.

The UK Millennium Cohort Study (Pereyra-Elías et al., 2022) followed 7,855 singletons to age 14. Longer breastfeeding was associated with higher verbal and spatial scores. Adjustment for socioeconomic position roughly halved the effect; further adjustment for maternal cognitive scores removed the association at age 5 but not at ages 7, 11, and 14. At 14, breastfeeding for 12 months or more versus never was associated with a 0.26 standard deviation difference in verbal scores (0.18 to 0.34), roughly 4 IQ points. That is a modest, persistent association — not a transformative one.

A 2025 ALSPAC analysis examined 373 neurocognitive outcomes and found 42 adjusted associations at p < 0.001, including full-scale IQ differences of +4.11 points at age 8 and +5.12 at age 15. The study is large and rich, but it is observational; the authors adjusted for social factors, not for every possible confounder, and a p < 0.001 threshold across 373 outcomes does not by itself correct for the number of tests. It is evidence that the association is robust, not that the effect is causal.

Mechanisms and competing explanations

Several mechanisms could produce a real effect. Breast milk contains DHA, ARA, cholesterol, oligosaccharides, and immune factors; it supports brain growth and may reduce infections that impair development. Nursing also involves skin-to-skin contact, responsive interaction, and hormonal effects that could influence maternal behavior. If the effect is nutritional, it should be larger in settings where formula is less adequate; if it is behavioral, it should be larger where nursing is part of a broader pattern of responsive care.

The confounding explanation is simpler. Breastfeeding is a marker of maternal resources: older, more educated, higher-income mothers with more partner support and less stressful lives are more likely to breastfeed, and those same mothers provide more cognitive stimulation. Adjusting for measured confounders reduces the association; sibling comparisons reduce it further; the randomized trial, which breaks the link between family background and feeding, finds a much smaller effect. That is the signature of confounding, not of a strong biological effect.

Reverse causation is possible in the other direction. Infants with neurological, oral-motor, or developmental difficulties may breastfeed less successfully, which would make the breastfed group healthier and could bias the association upward. Measurement error — recall of duration, crude categories, formula supplementation — usually attenuates effects. The net bias is uncertain, but the pattern across designs is consistent.

The exposure itself is heterogeneous. "Breastfeeding" can mean exclusive nursing for six months, mixed feeding, pumped milk, or a few weeks at the start. Studies rarely measure intensity, mode, or milk composition. A woman who breastfeeds for a year differs from one who stops at one month in ways that are not captured by a duration variable. This heterogeneity makes it hard to compare studies and easy to overgeneralize.

Finally, the outcome is heterogeneous. IQ tests measure verbal, spatial, memory, and processing speed; a global score can hide domain-specific effects. The PROBIT follow-up found verbal and memory effects but not performance. The sibling studies used achievement tests or vocabulary tests, not full batteries. A small verbal effect may be real even when a full-scale effect is not.

Evidence quality and verdict

Here is the synthesis I would defend.

The raw association is real and moderate. Breastfed children score about 3 to 7 points higher in observational studies, depending on the cohort and the adjustment. That is not a myth.

Most of the raw association is selection. The best sibling comparisons and the 16-year randomized follow-up reduce it to near zero or to a small verbal effect. The 2008 trial's full-scale estimate was too imprecise to support its conclusion; the 2018 follow-up is the more informative result.

A small causal effect is plausible. The Pelotas cohort, the Millennium Cohort, and the PROBIT instrumental-variable analysis point to a verbal effect of roughly 1.5 to 3.5 IQ points and a full-scale effect of roughly 0 to 3 points. The effect may be larger in settings where breastfeeding also protects against infections and poor nutrition, and smaller in contemporary high-income settings where formula is nutritionally adequate.

The effect is not a general intelligence boost. The domain-specific pattern — verbal and memory, not performance — argues against a broad cognitive acceleration. A 2- or 3-point verbal difference is meaningful at the population level and invisible in everyday life.

Calibrated confidence: ~75% that most of the raw association is explained by confounding and selection; ~60% that the true causal effect on full-scale IQ is between 0 and 3 points; ~25% that it is between 3 and 5 points; ~15% that it is larger than 5 points. I am ~85% confident that breastfeeding does not cause a large (>5 point) full-scale IQ gain, and ~65% confident that any causal effect is concentrated in verbal ability. These are not independent; the distribution is my summary of designs that disagree.

Put differently: the best estimate is a small effect, not zero, and much smaller than the popular claim. The confidence that the effect is not zero is lower than the confidence that the effect is not large.

What would change the answer

Five kinds of evidence would move this estimate.

First, a longer follow-up of PROBIT to age 30, with the same cognitive battery and better measurement of breastfeeding intensity. The 16-year results are the strongest causal evidence we have; a 30-year follow-up would test whether the verbal effect persists into adult outcomes.

Second, a prospective sibling study with measured maternal cognitive ability, home environment, and differential investment. Existing sibling studies are limited by achievement tests, small samples, and unmeasured between-sibling differences. A study designed to separate milk from mothering could resolve the positive Evenhouse result.

Third, a natural experiment that changes breastfeeding rates exogenously — maternity leave policy, marketing restrictions, hospital accreditation — with long-term cognitive follow-up. These designs avoid the self-selection problem without randomizing mothers.

Fourth, better genetic instruments. The current Mendelian randomization uses variants associated with a behavior; future instruments might target lactation biology, milk composition, or the child's metabolic response. A well-powered analysis with valid instruments could distinguish a small causal effect from zero.

Fifth, a preregistered replication of the FADS2 interaction with adequate power and a prespecified subgroup analysis. The current meta-analysis is a useful corrective, but the hypothesis is not dead; it is unresolved.

Until one of these arrives, the honest summary is the one parents and policymakers can use: breastfeeding has well-established short-term health benefits, and its cognitive benefit is probably small, mostly verbal, and not a reason to pressure mothers who cannot or do not breastfeed.

References

1. Kramer MS, Chalmers B, Hodnett ED, et al. Promotion of Breastfeeding Intervention Trial (PROBIT): a randomized trial in the Republic of Belarus. JAMA. 2001;285(4):413–420. doi:10.1001/jama.285.4.413. PMID 11242425. https://doi.org/10.1001/jama.285.4.413
2. Kramer MS, Aboud F, Mironova E, et al. Breastfeeding and child cognitive development: new evidence from a large randomized trial. Archives of General Psychiatry. 2008;65(5):578–584. doi:10.1001/archpsyc.65.5.578. PMID 18458209. https://doi.org/10.1001/archpsyc.65.5.578
3. Yang S, Martin RM, Oken E, et al. Breastfeeding during infancy and neurocognitive function in adolescence: 16-year follow-up of the PROBIT cluster-randomized trial. PLOS Medicine. 2018;15(4):e1002554. doi:10.1371/journal.pmed.1002554. PMID 29677187. https://doi.org/10.1371/journal.pmed.1002554
4. Der G, Batty GD, Deary IJ. Effect of breast feeding on intelligence in children: prospective study, sibling pairs analysis, and meta-analysis. BMJ. 2006;333(7575):945. doi:10.1136/bmj.38978.699583.55. PMID 17020911. https://doi.org/10.1136/bmj.38978.699583.55
5. Evenhouse E, Reilly S. Improved estimates of the benefits of breastfeeding using sibling comparisons to reduce selection bias. Health Services Research. 2005;40(6 Pt 1):1781–1802. doi:10.1111/j.1475-6773.2005.00453.x. PMID 16336548. https://doi.org/10.1111/j.1475-6773.2005.00453.x
6. Colen CG, Ramey DM. Is breast truly best? Estimating the effects of breastfeeding on long-term child health and wellbeing in the United States using sibling comparisons. Social Science & Medicine. 2014;109:55–65. doi:10.1016/j.socscimed.2014.01.027. PMID 24698713. https://doi.org/10.1016/j.socscimed.2014.01.027
7. Brion MJA, Lawlor DA, Matijasevich A, et al. What are the causal effects of breastfeeding on IQ, obesity and blood pressure? Evidence from comparing high-income with middle-income cohorts. International Journal of Epidemiology. 2011;40(3):670–680. doi:10.1093/ije/dyr020. PMID 21349903. https://doi.org/10.1093/ije/dyr020
8. Victora CG, Horta BL, Loret de Mola C, et al. Association between breastfeeding and intelligence, educational attainment, and income at 30 years of age: a prospective birth cohort study from Brazil. The Lancet Global Health. 2015;3(4):e199–e205. doi:10.1016/S2214-109X(15)70002-1. PMID 25794674. https://doi.org/10.1016/S2214-109X(15)70002-1
9. Horta BL, Loret de Mola C, Victora CG. Breastfeeding and intelligence: a systematic review and meta-analysis. Acta Paediatrica. 2015;104(467):14–19. doi:10.1111/apa.13139. PMID 26211556. https://doi.org/10.1111/apa.13139
10. Pereyra-Elías R, Quigley MA, Carson C. To what extent does confounding explain the association between breastfeeding duration and cognitive development up to age 14? Findings from the UK Millennium Cohort Study. PLOS ONE. 2022;17(5):e0267326. doi:10.1371/journal.pone.0267326. PMID 35613097. https://doi.org/10.1371/journal.pone.0267326
11. Caspi A, Williams B, Kim-Cohen J, et al. Moderation of breastfeeding effects on the IQ by genetic variation in fatty acid metabolism. Proceedings of the National Academy of Sciences. 2007;104(47):18860–18865. doi:10.1073/pnas.0704292104. PMID 17984066. https://doi.org/10.1073/pnas.0704292104
12. Hartwig FP, Davies NM, Horta BL, et al. Effect modification of FADS2 polymorphisms on the association between breastfeeding and intelligence: results from a collaborative meta-analysis. International Journal of Epidemiology. 2019;48(1):45–57. doi:10.1093/ije/dyy273. PMID 30541029. https://doi.org/10.1093/ije/dyy273
13. Lu Y, Yu X, Feng Z, et al. Causal effects of breastfeeding on childhood health outcomes: a Mendelian randomization analysis of mental and physical health. Medicine (Baltimore). 2025;104(45):e44888. doi:10.1097/MD.0000000000044888. PMID 41204498. https://doi.org/10.1097/MD.0000000000044888
14. Goulding N, Northstone K, Taylor CM, et al. Differences in neurocognitive development between children who had had no breast milk and those who had had breast milk for at least 6 months. Nutrients. 2025;17(17):2847. doi:10.3390/nu17172847. PMID 40944234. https://doi.org/10.3390/nu17172847
15. Patnode CD, Henrikson NB, Webber EM, et al. Breastfeeding and health outcomes for infants and children: a systematic review. Pediatrics. 2025;156(1). doi:10.1542/peds.2025-071516. PMID 40240318. https://doi.org/10.1542/peds.2025-071516
16. McGowan C, Bland R. The benefits of breastfeeding on child intelligence, behavior, and executive function: a review of recent evidence. Breastfeeding Medicine. 2023;18(3):172–187. doi:10.1089/bfm.2022.0192. PMID 36749962. https://doi.org/10.1089/bfm.2022.0192