While breastfeeding is of clear benefit to babies in terms of providing them with efficient and wholesome nutrition, the idea that it also stimulates cognitive and behavioural development does not currently enjoy robust empirical support. Nonetheless, this lack of empirical evidence hasn’t prevented some advocacy groups from continuing to make such assertions. Therefore, when I saw the headline in Monday’s Guardian – “Breastfeeding leads to better behaviour in children, researchers claim” – I immediately had a sense of foreboding.

First of all, determining the effects of breastfeeding on a child’s subsequent behavioural development is a devilishly difficult research challenge: for one thing, it is simply not the case that research methods used to establish biomedical realities (such as improved nutritional health) can be deployed in a straightforward manner to also examine behavioural outcomes (such as improved childhood behaviour). Secondly, previous studies have by and large been extremely ambiguous and often highly controversial, which does not inspire confidence. And thirdly, from a sociocultural perspective, arguments about breastfeeding seem very quickly to descend into arguments about morality and/or social class, in which bottle-feeding mothers are more or less accused of child neglect. Given these concerns, my own view is that researchers need to be extremely sure about their methods before making strong claims about non-nutritional benefits of breastfeeding. As several methodological problems plague this area of research, all of which seem to emerge again and again and again, a simple question arose. Would this study be any different?

The Guardian article encapsulates the study’s conclusion in a simple opening sentence: “Babies who are breastfed are less likely to have behavioural problems by the age of five than those given formula milk.” The article goes on to explain that the findings were based on data from a large-scale prospective cohort study, in which 4% of breastfed babies manifested behavioural problems in childhood, compared to 16% of bottle-fed babies. Behavioural problems, we are told, included things like “clinginess and anxiety, hyperactivity such as restlessness, and conduct problems such as lying and stealing.” The original research report is published in latest issue of the academic journal Archives of Diseases in Childhood, where further details about the study can be found.

Before reading the report in detail, I decided to conduct a brief academic exercise. I mentally prepared a list of possible research pitfalls that might adversely affect such a study’s validity. In no particular order, I came up with the following:

1. Failing to account for the role of maternal IQ. Many previous researchers point out that relationships between breastfeeding and behaviour might be confounded by other variables, such as education level. In other words, mothers who are, say, better educated may be (a) more likely to breastfeed and (b) better at curtailing their child’s misbehaviour. This would undermine claims that breastfeeding is the cause of good behaviour, because both breastfeeding and behaviour would be caused by something else. Several possible confounds have been studied to date (such as socioeconomic status, maternal age, maternal smoking, premature birth, etc.). However, one factor — maternal IQ — has been shown to be particularly important, but also, crucially, to be very often overlooked. In fact, reviews of multiple studies show that maternal IQ is one of the strongest confounding variables around, much stronger than, say, maternal age or socioeconomic status. When maternal IQ is accounted for, previously reported effects for breastfeeding on cognitive functioning disappear. Given that cognitive functioning typically predicts behavioural problems, researchers intending to examine links between breastfeeding and behaviour should really be sure to control for maternal IQ. A study that fails to account for maternal IQ cannot be considered conclusive.

2. Failing to account for maternal attitude toward child-rearing. Given the publicity around breastfeeding, and the well established point that breastfeeding is of nutritional benefit to babies, we might expect that progressive attitudes towards childrearing will count as another confounder. Such a link between breastfeeding and parenting attitudes has been demonstrated in previous research. This could well create a statistical association between breastfeeding and behaviour that is coincidental in nature, rather than causal. Therefore, a good study should examine, and account for, maternal parenting attitudes.

3. Failing to account for heritability of personality. Maybe how we treat children has no real effect on their behaviour. Instead, maybe it comes down to genetics. This would help explain why parents and children often appear to have similar personalities. One strongly inherited personality trait is calmness: parents with a calm disposition tend to have children who also have calm dispositions. As such, a calm parent may be (a) more likely to breastfeed and (b) more likely to have a child who is also calm, and thus less likely to develop behavioural problems. In order to rule out such confounds, researchers should statistically control for measures of relevant personality traits.

4. Failing to account for neonatal behaviour as a predictor of childhood behaviour. The prospective design of the study presupposes that establishing breastfeeding first and behavioural problems later will account for the reverse-causality problem. In other words, because behavioural problems emerge after breastfeeding, this means they cannot be considered causes of breastfeeding. However, this is insufficient because behavioural problems in childhood may be preceded by corresponding neonatal behavioural issues, which will certainly precede the choice to breastfeed. Many mothers attempt to breastfeed but fail because their babies just don’t take to it. Furthermore, we know that even newborn babies vary greatly in their sense of calmness, and we know that this neonatal disposition remains consistent into childhood. For example, studies have shown that children who show distress in the labour ward or during the first few weeks of life tend to have higher physiological reactivity to stress in later childhood. As such, boisterous or temperamental babies may be (a) less likely to be breastfed and (b) more likely to develop behavioural problems later. Therefore, when attempting to draw scientific conclusions about breastfeeding and childhood behaviour, neonatal dispositions should be somehow assessed and accounted for.

5. Failure to remove that subset of parents whose child had a clinical condition that prevented breastfeeding. Some children cannot be breastfed because they have a clinical condition (such as a developmental disorder) that also precipitates challenging behaviour in childhood. This is admittedly a very small number, but a thorough research study would at least classify these children and account for them in the analyses.

6. Failure to ensure reasonable comparisons. Many studies generate unreasonable comparisons. Mothers are put into one of two categories – breastfeeding or bottle-feeding – without recognising the fact that there are many variations within these categories (What if a mother breastfeeds for one year and bottle-feeds thereafter? What if a mother expresses breast milk that a father then gives to the baby via a bottle?). A related issue is whether the researchers have succeeded in establishing a dose-response effect (in other words, whether the amount or duration of breastfeeding relates directly to the amount of benefit to the child). If there is a meaningful benefit to be derived from breastfeeding, then a dose-response effect should be evident; if a dose-response effect is not present, then the link is likely to be coincidental.

7. Failure to validate measures. Many such studies rely on retrospective self-reporting to establish whether breastfeeding took place, and many rely on parental self-report to measure behavioural or psychological variables. Use of self-report for the latter can be particularly unreliable, especially if parents are being asked to describe their children’s bad behaviour. Some parents will downplay their children’s behavioural problems due to social desirability bias; others may exaggerate them as a way of highlighting the burdens they have to face in life. As there is no way to be sure, a good study should use an objective means of determining the nature and extent of children’s behavioural problems. Several standardized methods exist. This is crucial to the soundness of the study overall, as everything hangs on whether we can be sure which children are genuinely showing behavioural problems and which are not.

I’m afraid I won’t need to exhaust you with too much detail about what comes next. This is because, when I read the original research report, my sense of foreboding was confirmed. Quite simply, none of the pitfalls I predicted beforehand were properly avoided in this study.

Maternal IQ was not measured (only “mother’s education” came near, but this is not the same as maternal IQ and does not have the same relationship with breastfeeding). Nor was maternal attitude to childrearing, maternal personality, or neonatal behaviour. Babies born extremely prematurely were systematically excluded because of likely feeding complications, but babies with other conditions that complicate feeding were not systematically excluded. The comparison included in the study was at the simplest grain of analysis; namely, breast-feeders versus bottle-feeders, without accounting for variations within these categories. There was no dose-response relationship. And finally, behavioural problems were assessed using parental self-report only, with no method of triangulation (against, say, teacher or physician reports) or other forms of objective verification. However, in fairness, the questionnaire used by the researchers was previously validated as a reliable measure in prior studies. But overall, this is hardly a conclusive study. More research is certainly still needed — so maybe the next study will be the one that avoids these recurring pitfalls (although forgive me if I don’t hold my breath).

Interestingly, when discussing their findings, the researchers elaborated well beyond the dataset in order to provide some theoretical speculations as to why breastfeeding might lead to better behaviour in children. This mainly comprised two very old, but as yet unsubstantiated, arguments: firstly, that breast milk contains nutrients that somehow lead to enhanced childhood behaviour; and secondly, that it is the the intimate contact between mother and child which produces positive outcomes. However, neither of these mechanisms was tested in any way by this study and so their inclusion in the report might be seen as superfluous — or even oratorical. But a big problem is that these explanatory theories did find their way into the media reports on the research study (as well as the Guardian, see for example the BBC News). In all likelihood, this would have given many readers the impression that the study actually tackled these theories, when all it really found (or even looked for) was a questionable statistical association between breastfeeding status and parental reports of child behaviour. In reality, these two theories, and the link between breastfeeding and cognitive or behavioural outcomes itself, remain highly controversial.

It must be repeated that, without doubt, breastfeeding promotes the physical well-being of infants. As I mentioned in a previous post, there appears to be ample justification for breastfeeding without having to make unwarranted claims about its impact on cognitive or behavioural outcomes. In fact these types of assertions can turn out to be counter-productive. Making such claims in the absence of reliable evidence weakens the overall position of breastfeeding advocates, because it exposes them to the accusation that they are exaggerating their case. Furthermore, given that breastfeeding is less common among mothers from low socioeconomic backgrounds, such claims may perpetuate stigmatizing class stereotypes.

It would be especially ironic if that was the outcome of the fashionable left-wing cause to promote breastfeeding — the reinforcement of right-wing slurs about the lower classes, with their unruly and unintelligent kids.

Brian Hughes

Brian Hughes is an academic psychologist and university professor in Galway, Ireland, specialising in stress, health, and the application of psychology to social issues. He writes widely on the psychology of empiricism and of empirically disputable claims, especially as they pertain to science, health, medicine, and politics.