Breastfeeding Doesn't Affect Intelligence

There is a lot of research on breastfeeding and IQ, and yet a huge amount of it is very poor. Many of them make no effort to control for parental characteristics, and yet we know that, because of the prevailing view that breastfeeding is important for intellectual and other development, that parental intelligence predicts breastfeeding, making it difficult to ascertain whether breastfeeding helps intelligence or whether just intelligent families breastfeed. In this piece, I will give an example of research that has been used to overstate the case for the importance of breastfeeding in cognitive development, before then showing that the effect of breastfeeding on intelligence is probably in the range of -1 to 1 IQ points.

Before diving in, I should probably point out that 1 IQ point sounds tiny and it certainly isn’t noticeable between individuals, but it isn’t nothing. Suppose that you had 5 interventions, of which breastfeeding might be one, that each produced a 1 IQ point gain additively, then that would accumulate to a 1/3 of a standard deviation increase, which could take a kid from performing 15th in a class of 30 to performing 11th best, or a kid performing 5th best in the class to performing 3rd best. That isn’t nothing! But, I am showing that at best breastfeeding could be part of an overall strategy of interventions to try to raise a child’s intellectual performance in future, and is not by itself sufficient to achieve any noticeable gains over bottle feeding and keeping everything else constant.

This matters because beliefs about breastfeeding affect people’s decisions! Some mothers really love breastfeeding. Others don’t love love it, but nonetheless gain satisfaction from the bonding and caring experience with their baby. But there are women who find it a chore or otherwise have difficulty with it, and it likely does play a role in deciding whether to have an additional child for some women. I currently have half of a paper that I need to finish in which I show that a baby formula scare in Israel led to lower fertility rates amongst the demographics that relied on it the most, showing that people are indeed sensitive to the perceived ability to use an alternative option to breastfeeding.

Anyhow, here is my example study. I find it an interesting one on which we can zoom in, because it does make the appearance of being rigorous. It uses socioeconomic status controls, and it tries to investigate the mechanism of whether breastfeeding affects brain and neural volume, which in turn predict IQ. But there are limitations. The sample was just 50 in total, and they found no statistically significant relationship in the total population. In the sample of 24 girls, they also found no significant effects. But in the sample of 26 boys, they found an effect. This study was then reported in news outlets as proof of the positive effects of breastfeeding on IQ, all while it failed to find any effect for girls! The study also accepted in limitations that it didn’t fully control for parental IQ.

Of course, it could well be true that breast milk only matters for boys’ cognitive development (which is a pretty interesting possibility itself), but it is very striking all the same that a study would be used to perpetuate a view (‘breastfeeding always good for IQ’) which it clearly does not support (e.g., ‘breastfeeding good for boys’ IQ, neutral for girls’ IQ’). It also short cuts interesting implications! Boys have higher variance in the speed of their cognitive development and generally develop their academic abilities a bit behind the curve compared to girls. My personal speculation on the explanation for the male-female disparity is that breastfeeding likely incurs an advantage in the speed of development (the study tested cognitive performance whilst the kids were teenagers), rather than adult intelligence. Also note that the benefits for the lads were not sufficient to make the population average effect statistically significant, meaning that if the paper had never divided things up by sex, it would have instead been reported as failing to find significant effects (if it still ever got published—positive result bias and all).

Anyhow, here is the study that I find really interesting. Ultimately, methodology matters, and if 30 studies out there conclude different but all fail to adequately control for confounding variables, then I pick the one without the confounding over the 30. Whenever studies try to control for parental characteristics, it’s difficult to do it completely. I might control for parental social class, but it might still be true that, within each class, the more intelligent parents are also more likely to do breastfeeding. I might then try to control for maternal IQ, but even then I’m missing a very important control: paternal IQ. This stuff matters! Examination on the effects of IVF on rates of detectable birth defects eventually identified that nearly all of the raised risk went away once paternal age was controlled for, leaving a statistically non-significant result for the estimated effect. In the end, the age of both parents is the big risk factor, not IVF. But many still believe that the father’s age doesn’t matter and that IVF causes detectable birth defects.

The best study of which I’m aware which tries to properly account for confounding variables on breastfeeding compares siblings. Full siblings share both parents and they share a home environment and a locality, in addition to sharing the basic socioeconomic background characteristics on account of… having the same parents. When we look at siblings, non-breastfed children actually outperformed their breastfed siblings on some cognitive tests, although by statistically insignificant amounts. Now, I’m not really suggesting that breastfeeding harms IQ relative to bottle feeding: the result was statistically insignificant by a long way and the estimated magnitude was pretty small anyway (-0.63 points, or 0.042 standard deviations). But for the finding that, with controls, breastfeeding raises IQ by maybe between 0 and 1 IQ points and the fact that the sibling differences study suggests a negative effect on IQ in the region of 0 to -1, this is indicative that the full ‘reasonably likely’ range is -1 to 1. This is, in my view, an important distinction from estimates that the IQ benefit might just be, say, a statistically non-significant 0.5 IQ points, because there isn’t even consistency on the directionality of the effects.

You might worry about the fact that breastfeeding variation between siblings tends to be non-random, and I agree that they made a very strong methodological assumption there. There’s one reason to expect the confounding from that to go in the other direction and inflate our estimates of the IQ effects from breastfeeding. Weaker babies are sometimes given the bottle because it’s easier for them—we would intuitively expect such babies to be more likely to have lower intelligence than their healthier babies, because general health and intelligence are (weakly) correlated. But, it’s also true that mothers tend to breastfeed later-born children more than earlier-born babies (often because they’ve got more familiar with it). We also know that birth order weakly negatively predicts intelligence, perhaps because of reduced parental attention and almost certainly at least in small part because of a higher mutational load as parents age, as I’ve written before. Given we are dealing with small estimates, it’s entirely possible that this confounding from birth order could turn something like a 0.3 IQ point boost to a 0.6 IQ point knock downwards in their estimates. But, because the correlation with IQ and birth order is small (something like a 1 or 1.5 IQ point drop for every rank one drops in the birth order) and the correlation between birth order and expected breastfeeding duration isn’t crazy long (something like 2 months), then the confounding with sibling breastfeeding rates is very unlikely to be turning a large positive effect from a negative one.

And there’s a way we check this: the same study with the sibling comparisons also ran a regression model that tried to control for as many confounders as possible, including birth order. It found an estimated positive effect on IQ from breastfeeding of… 0.52 IQ points. The result was not statistically significant. Controls did not include paternal IQ or education.

So, the sibling method is likely to be a negatively biased estimate of the effect of breastfeeding on IQ due to the confounding of birth order. The multivariate regression is likely to be a positively biased estimate of the effect of breastfeeding on IQ since it still can’t get controls on paternal characteristics, which are likely to affect the chances of breastfeeding. One estimate was about -0.6, the other was about +0.5. Add in a bit of error, and we’re looking at an IQ effect somewhere from -1 to +1. But it’s more likely that it lies in between -0.6 and +0.5. There’s a number I’m thinking of that is pretty much squarely between -0.6 and +0.5: the number is zero. It seems to me at this point that breastfeeding, if it does affect IQ, does so negligibly.

Regardless, as best I can tell ‘breast is best’ in as much as it brings other health benefits. But the idea that it is crucially important for IQ doesn’t survive contact with attempts to better control for confounding variables.

This is important to get right, because it is likely that for some women the belief that breastfeeding is an essential for doing motherhood properly becomes a cost that contributes to choosing not to have an extra kid. And for some mothers who are struggling with it, it will be having a deleterious effect on their experience that does not reflect the trade offs they would be willing to make given the best evidence on this.

---

Some claim that we should throw out this research because clearly something that is natural and so long-standing cannot not offer benefits above a synthetic alternative. Kristin Lawless made this argument just 3 days ago to buttress a different argument made by Emily A. Hancock. I have limited time for such thinking personally.

For one, we have evolved significantly just in the past ~12,000 years of moving from hunter gatherers to agriculturalists. We are literally genetically distinct on average from our ancestors of just 12,000 years ago. To give two examples, we are smarter and we can digest more stuff. Our genetic predisposition for educational attainment also rose by faster than our genetic predisposition to develop cognitive skills, suggesting that we also changed in other traits. That makes me a bit less tempted to automatically go ‘oh, well, if our hunter gatherer ancestors did it then it must be the enlightened thing’. Our hunter gatherer ancestors also lived in poverty and had homicide rates that make Americans look like civilised people. Maybe we do some things better today. Granted, this doesn’t mean this argument never implies useful conclusions. But it does mean that we shouldn’t rely on ‘our hunter gatherer ancestors did it’ or ‘it was the way’ as our argument for rejecting things. And, of course, I’ve refrained from taking cheaper shots about the naturalness of opium and the non-naturalness of modern antibiotics (although now I haven’t kept up the refraining, but I make no apologies on this).

The argument from Emily Hancock against downplaying benefits of breastfeeding is more refined. She makes the case that it is part of what it is for us to be human, not only because of the length of time we’ve been doing it, but also because of our biology today. When mothers are in late pregnancy, their bodies prepare to lactate. When mothers give birth, they lactate. This is part of womanhood. And I take her point. And breastfeeding is often about more than health outcomes—as I said at the beginning of this, many women gain a lot of satisfaction from it and bonding with their baby, both of which are super super great things. It’s also crazy cool to see your baby simultaneously hug his or her mother as closely as humanly possible and get fed at the same time, particularly when their eyes also look drugged up with the hit of the sweet milk. I don’t renounce these things, nor am I trying to attack breast milk or break up the mother-baby relationship.

But I do believe in trying to get things as accurate as possible and spreading that, so that each woman can work out what works for her. And I worry that the alternative could create needless distress and even result in a handful of fewer people coming into the world who would have been very much loved and raised with care.