Skip to content

Is it really true that babies should sleep on their backs?

Asher Meir writes:

Arnold Kling is a well-regarded economics blogger. Here he expresses skepticism about the strength of the evidence behind recommending that babies sleep on their backs.

I recall seeing another blogger expressing the same doubt at some length, or maybe it is another post by Arnold, I can’t find it right now.

Of course there are many cases of leading experts spouting authoritative advice with little evidence, but this is in many ways more worrisome. First of all they are claiming that ignoring their advice is life-threatening which is quite coercive. In fact I saw at least one court case where a mother was held liable in part for not putting her child to sleep on its back. In addition, there is really little harm to avoiding eggs even if they are not really bad for you, but the SIDS advice is affecting the well-being of millions of infants around the world as well as their parents. If the advice is not necessary it is extremely harmful just because of the lost sleep of parents and babies, and it is at least plausible that the advice is actually harming babies in other ways as Arnold mentions.

Obviously if the research is carefully done then the physicians have an obligation to warn parents not to endanger their babies but I think it is worth looking into the strength of the evidence on this.

Meir summarizes:

I am worried that this may be a redux of the ridiculous unfounded nutritional advice we have been getting for years, except that the nutritional advice was mostly merely silly and harmless and this sounds potentially quite harmful. But who knows. Let’s let your readership weigh in.

I have no idea, myself. We did follow this advice with our kids, but I’ve never looked into the statistics on it; I just assumed it was a good idea to follow the general recommendations.


  1. Adam says:

    I followed the links, but but didn’t find any evidence. Arnold Kling just claims things. Why should I trust an economics blogger (however well-regarded) to know anything about medicine?

    • Anoneuoid says:

      This article goes into it:

      It sounds like the usual pattern, a correlation is seen between a poorly defined medical issue and some everyday phenomenon and then the public health advocates run with it without any regard for cost-benefit analysis, etc. Then the definitions change, events are moved from the target category to some other category, and success is claimed.

      Other people are thinking that babies don’t sleep as well on their backs, which is leading to autism:

      • Brian says:

        When your sources are the “World Scientific Journal” (published by renowned spammers Hindawi) and the “World Journal of Clinical Pediatrics” (published by Baishideng Publishing Group, on Beall’s now defunct list of predatory journals), then maybe you’re dealing with fringe arguments…

        • Anoneuoid says:

          Sorry, I just followed a link from those comments and searched pubmed. I have no idea about Hindawi, etc or how they could possibly be any worse than eg Elsevier. The entire academic publishing industry is a giant scam that cant get replaced soon enough, and I don’t see how someone could know about “Baishideng Publishing Group” without realizing that, so this post is kind of funny to me.

          Anyway, in general I don’t use the “journal heuristic” since I’ve seen so much crap in “prestigious” journals, and can judge the claims for myself. As long as the authors use proper citations and such its good enough for me.

          Do you have a problem with any of the content of those articles? I just skimmed them and they seemed straightforward, pretty much exactly what I would expect. I by no means stand by them or vouch for them though.

          • Brian says:

            Fair enough, but I’m not trying to make a point about journal prestige per se. The issue is that there are a growing number of journals that are *literally* scams, in the sense that they claim to provide peer review of articles, but do no such thing. I’ve lost count of the number of emails from Hindawi journals inviting me to submit an article for a given issue that is due to be published in 10 days (naturally on a topic that I have no knowledge or interest in). They are mostly outlets for cranks to publish their ludicrous ideas.

            I appreciate that you should judge articles on their individual merit, but life is too short, and I just don’t have that much interest in the topic.

    • Noah Motion says:

      Why should I trust an economics blogger (however well-regarded) to know anything about medicine?

      I think a fairly high level of general skepticism is warranted, given how much we know about questionable research practices and the weakness of a lot of evidence for claims in a wide variety of fields. Kling’s positive claims:

      I am convinced instead that it leads to widespread sleep problems among babies at least until age four, it slows their motor development and probably their cognitive development, and leads to many children wearing helmets to reshape their heads.

      may or may not have strong evidence backing them up, and so also deserve some skepticism. But his main negative claim there:

      I don’t find convincing the evidence that this reduces death of infants.

      seems pretty reasonable to me. If I don’t have the time to dig into the research, it seems reasonable to imbue a smart person like Kling with some trust (though my uncertainty with respect to this is quite high), and this kind of skepticism seems like a reasonable default position to hold.

  2. JFA says:

    Yeah… Kling’s post had no evidence and just expressed skepticism and claimed babies sleeping on their back just led to sleep problems and didn’t do anything to reduce SIDS deaths (again no evidence presented. Here’s a chart ( from this NPR article ( I certainly wouldn’t compare it to nutritional advice, as the recommendation to avoid fat coincided with an increase in obesity rates. At least in the case of SIDS, the advice seems to be associated with good rather than bad outcomes. I don’t know if it’s causal. Lot’s of statistics for bad things (rates of various crime, teen pregnancy, etc.) show declines starting in the early 90s. But I will say, given the expected cost of being wrong and the absence of data that says back sleeping prevents infant death, I’d make sure my kids slept on their back.

  3. Thanatos Savehn says:

    I strongly disagree with this claim: “… the nutritional advice was mostly merely silly and harmless …” Poor inferences brought us epidemics of food allergies, cardiovascular disease and obesity. If you let an idiot design your product and someone gets hurt by it you get sued. If you use your credentials to peddle ideas that get people hurt you just go ¯\_(ツ)_/¯ and move on to designing and marketing your next fashionably stupid inference.

    P.S. We also followed the advice with our kids even though the evidence for a protective effect of back-sleeping was weak at best because … well because of the same reason we had cameras with microphones pointed at them and still rushed upstairs to check on them any time we couldn’t hear them breathing.

  4. Steve says:

    JFA’s point is well stated. We cannot do experiments on infants. We are stuck with observational data and probably studies that rely on parents memories of whether they put their infants to sleep on their stomachs. Of course, those studies are going to be suspect. But, public health officials have a serious crisis to deal with and have to make a recommendation based on uncertain data. SIDS is rare, and sleeping on the stomach is common. So, of course, the one is not a sufficient cause of the other but before the actual mechanisms can be established, back-sleeping is the only intervention that seems to reduce the risk. The level of certainty that we seek has to be adjusted to the outcomes we are responding to.

    • Dzhaughn says:

      There was no crisis, except an epidemic of media coverage.

      • JFA says:

        Here is a list of case fatality rates of various diseases: You’ll see that malaria, hepatitis A, and various influenzas have/had a death rate of around 1 per 1000 diagnosed. In 1988, the death rate from SIDS was around 1.4 per 1000 live births. So here we have a population that has the condition of being an infant and around 1.4 per 1000 died because they they happen to be infants. I don’t know if that should be considered that a crisis, but it seems a bit blithe to suggest the only problem was media coverage.

    • Noah Motion says:

      back-sleeping is the only intervention that seems to reduce the risk.

      Based on the link that Nick Matzke posted below, and based on some of the papers I found poking around on this topic yesterday, I don’t think this is true. Having infants sleep on firm bedding, having them sleep in the same room (but not the same bed) as the parents, preventing over-heating, breast-feeding, and having the parents not smoke are also given as recommendations on par with back sleeping, I assume because of similar relative risks or odds ratios estimated from observational studies.

  5. I think Moms tend to position babies in a crib so that they sleep on their tummies. It’s easier to lay them down that way.

  6. Noah Motion says:

    Here’s an reasonably solid looking study on sleep position and motor development. I haven’t looked at it in detail, so I don’t want to overstate how strong the evidence is, but they claim that supine sleep is associated with delayed (relative to prone sleep) motor milestones, but that all of the infants reached motor milestones within the accepted normal developmental range. The last line in the abstract is “This difference in milestone attainment is not a reason to abandon the American Academy of Pediatrics’ sleep position recommendations.”

    • Noah Motion says:

      At this rate, I’ll be an expert in this field by the end of the day! Here’s a study of sleep position and SIDS with total N = 3,110, but with only 23 instances of SIDS, and with only 15 of those with sleep position data available. The key odds ratio is 4.47, with 95% CI = [1.30, 15.43]. Not nothing, but not terribly compelling, either.

  7. oncodoc says:

    I agree with JFA also. There are commentors on the Kling site that support his position by saying that there were no sleep position recommendations in the good old days, and we survived. Survivorship bias remains alive! Additionally, the SIDS mortality even before the last decade was less than 2/1000; in a pretechnological world with infant mortality rates of 50/1000, SIDS might have escaped notice.

    • Elin says:

      Actually there were recommendations and they were specifically for babies to sleep prone. Perhaps some of those commenters left the interaction with nurses and pediatricians in the maternity ward to their spouses. Or maybe they are too young to know. Either way, always better to do a short literature search yourself before posting something incorrect.

  8. In so far as the commentary on nutritional advice, I think the quality of nutrition intake makes a difference. I don’t know why nutritional experimentation should be disparaged. Maybe the studies were too limited.

    • I don’t think experimentation is so much the issue, it’s the observational junk combined with weak theory that leads to “eating butter is bad for you, you should switch to margarine” and then 2 decades later “whoops, margarine is worse than butter due to trans fats, go back to butter” etc.

      No one should dispute that high quality nutrition leads to better health outcomes than low quality nutrition. The bigger issue is among first world countries, what kinds of marginal benefit can be had by tuning the grossly OK nutritional intake of a typical say middle class American, and what specific recommendations are really warranted by causal data?

      For example, should people with average american diets consume more nuts for cardioprotective effect, or is that not really a thing? Does reduction of animal fat intake in middle class american diet have any real effect? Did it have an effect in the past but no longer? Does the effect depend on what the calories are replaced by? (answer is obviously yes, but exactly how does that work?)

      The tendency to do some smallish usually observational study or maybe RCT but primarily looking at intermediate outcomes that aren’t very important themselves (say cholesterol levels for example) and then find a p less than 0.05 result and declare PROBLEM SOLVED leads to overall pretty lousy knowledge in this field.

  9. Eric says:

    The campaign to put babies to sleep on their backs had an almost immediate and very dramatic effect. It saves the life of about 1 baby per 1,000-2,000 births in the US. That amounts to several thousand lives saved per year in the US alone. You can see the overall trends here:, but this conclusion is documented through lots of studies. There are likely to be some people reading this blog who wouldn’t be here if not for that campaign.

    I’m not sure what’s more distressing in this thread – the original misinformed blogger, or the responses in this thread from people who should know from reading this blog that it’s really unwise to try to draw conclusions by looking up single studies in Pubmed.

  10. Nick Adams says:

    Yes, I have a big problem with the content of those articles.
    Try something a bit more sensible –
    Non-expert opinion versus a good observational study is a rather uneven contest.
    The study was conducted in Australia though, which may not exist…

    • Anoneuoid says:

      Yes, I have a big problem with the content of those articles.

      I suspect this was supposed to be a response to my comment above… if so, what is the big problem that you have? Otherwise, please ignore this post.

  11. Nick Matzke says:

    The discussion in the thread is a bit odd, it’s not like there are no obvious mechanisms available for why sleeping-on-the-back might be better. The main mechanism is that tummy sleeping might make breathing more difficult / increase the chance of bedding interfering with breathing / reduce the infant’s ability to throw off bedding that covers the face, makes them too hot, etc. Most babies would probably be fine either way, but perhaps on occasion, for a baby with a weak startle reflex, low birthweight and low strength/energy, other breathing issue (having had a cold, smoker in the house, etc.), sleeping on the back makes the key difference.

    Given plausible mechanisms, I think it is reasonable to base action even on only moderately strong statistical results, especially once it seems likely that a controlled study would cause risk to tummy-sleeping babies. (Although, my sense was that the statistical evidence was pretty good also, but I haven’t dug into it.)

    (Conflict of interest statement: I am currently watching my 6-day old newborn sleeping on her back, wrapped snug but not too tight, in natural-not-synthetic fiber clothes, feet at the bottom of the bassinet so she can “kick out” if distressed, etc.) (And she sleeps like the dead in basically any position after she’s been fed.)

  12. Nick Matzke says:

    PS useful link, just to summarize what the medical recommendations are, and the diverse factors that correlate with SIDS:

  13. BenK says:

    The nutritional advice was far from silly and harmless. People traded fat for sugar, salt, and starch, and it killed them.
    They traded good milk for swill like skim, with stabilizers and unbound sugars.
    They traded not great for you sweets for artificial sweeteners and … got C. difficile, among other things.

    That advice about eggs was bad. Trying to eliminate cholesterol from the diet masked real struggles with inflammation and infection.

  14. Ronan says:

    Yes, they should.

  15. tomfinucane says:

    Dr. Spock taught the prone position. People learned that.
    In 1993, “Last year’s American recommendation (back to sleep) is controversial among the pediatricians who believe that the prone position helps reduce the risk of inhaling stomach fluid.
    Who’s giving odds on what recs will look like in another 25 years?

    Nutrition advice upregulates credulity in humans.

  16. Elin says:

    I think the main advice to Kling and Meir should be that if they are skeptical they should conduct a literature review at least at the level that we’d expect of a college junior in nursing or public health. These days you don’t even need to be a faculty member or in a major city to search the scholarly literature and pretty often you will find copies of the full articles if you spend a little time looking. Further if at a university another option is that they could walk across campus and speak to faculty who are knowledgable in this area. Skepticism is generally a good thing, but turning that into an argument about a topic where you have no expertise and then publishing it as though you do is really borderline academic malpractice.

  17. Joshua Pritikin says:

    To interject some evidence into the discussion, cow’s milk is suspected to increase the chance of SIDS. See and

  18. markus says:

    FWIW, when I looked into the matter a while ago, the back sleeping recommendation was supported by multiple studies and is basically sound.
    There are IMO two unrelated problems with the research. The first is due to the observational and post-hoc nature of the studies which cause factors such as ‘smoking’ to pop up as risk factors and it’s wholly unclear whether that’s encoding lower socioeconomic status, more risk taking in health matters or whether researchers seriously want to claim that having a single cigarette after lunch break endangers one’s babies. Similarly, ‘baby wasn’t sleeping on back’ in the case of a SIDS case may encode parents not caring about the latest advice (with attendant lower regard for health and safety, fewer check-ups etc.), a fussy baby, etc. My overall impression from the research is that researchers try to account for as many of these alternative explanations as possible and the consistency of results in idfferent samples and with different control variables suggests that the main recommendations are based on having successfully indentified causal factors.
    The second problem is that recommendations and the communication on them have to sound more certain than the underlying science warrants.

    • Anoneuoid says:

      What does it mean for the research to be “basically sound”? I posted this which seems (on the surface) to be a reasonable assessment of the evidence:

      SIDS mortality data derive exclusively from epidemiological studies and both the AAP recommendation and the BTSC were launched primarily on the basis of such reports from various countries outside the U.S. However, while it is axiomatic that the validity of any epidemiological study is critically dependent on diagnostic accuracy, there exists today no global consensus on a definition of SIDS. For example, Sawaguchi et al.[5] acknowledge that “a SIDS diagnosis is not uniform throughout Japan and such a diagnosis is not made based on any internationally recognized definition.” Official views concerning a SIDS diagnosis in Japan differ among pediatricians, legal scholars of forensic medicine, and pathologists. In the same reference, they note that in Scotland in the 1980s, many cases of sudden unexpected infant death (SUD) were reported as “SIDS”, whereas in the 1990s there was a trend reversal and the majority of similar cases were classified as “unknown”.

      In 1999, 5 years after the start of the BTSC, Cote et al.[6], in a study of 623 cases of SUD in infancy, found that the percentage of non-SIDS diagnoses was much higher for autopsies performed in a center with expertise in pediatric pathology than in a general hospital or medico-legal institute. Non-SIDS diagnoses were also much higher at age ranges atypical (12–18 months) as compared with typical (1–6 months) for SIDS.


      In 1991, an expert panel convened by the U.S. National Institute of Child Health and Human Development (NICHD)[11] expanded the criteria for establishing a SIDS diagnosis to include a thorough coroner’s investigation of the death scene and a review of the infant’s clinical history in addition to the performance of a complete autopsy. This was a critical change, because cases of suffocation due to unsafe sleeping environments or to infanticide could not be distinguished from SIDS by autopsy alone.


      By the time the BTSC was begun in 1994, the prevalence rate that year had already dropped to 1.03 SIDS deaths per 1,000 live births from the 1991 rate of 1.3 SIDS deaths per 1,000 live births.

      So the first claim is there are multiple reasons to expect a decrease of SIDs around that time. It is wrong to attribute the entire decrease to prone sleeping.

      Second, there are possible negative effects of prone sleeping that were never really considered:

      SIDS is a rare disease that claims the lives of approximately 0.5–3.0 infants per 1,000 births yearly in the developed countries of the world. By contrast, the incidence of positional facial and head deformities due to the supine position has increased dramatically since the BTSC and now affects 1 in every 60 live births[28].


      There are indications of a rapidly growing population of infants who show developmental abnormalities as a result of prolonged exposure to the supine position[29,30].


      It is likely that prone is the normal sleeping position for infants[33] and offers the deepest, most restful sleep[34,35,36,37]. Because infants spend more time in sleep than children or adults, it is logical to assume that sleep is even more important for their rapidly developing nervous systems and for preserving the integrity of their sleep cycles[38].

      Some other posters had a problem with these claims because of the publisher and for reasons they declined to mention.

      To me “sound science” would be that the researchers predicted a certain curve for what is seen in Fig 1 of that paper, ie “SIDs rate” vs “percent side and back sleeping” beforehand that took into account the effects of other changes like to the definitions, etc. Then when the data came in over the next decade they would compare the predicted effect vs actual. If the two are close, they are probably on to something.

      Second, sound science would not only focus on the benefits of a treatment/intervention but also have some accounting of the negative side effects and costs. This should be just as important as investigating the benefits and not superficial.

      So I dont see any particular reason to take this research seriously. I suspect it is another of those 30 year fads and we will soon learn that letting the baby sleep on the side it prefers and monitoring it closely is best. Or even that ideally babies should never even be put down for the first few months, instead they should always be held by family/friends and sleep in someones arms/lap.

      • Anonymous says:

        Some typos:

        These should say “supine” (on the back). Although the first may still be ok, the second is definitely wrong:
        “It is wrong to attribute the entire decrease to prone sleeping.”
        “Second, there are possible negative effects of prone sleeping”

    • > whether that’s encoding lower socioeconomic status, more risk taking in health matters or whether researchers seriously want to claim that having a single cigarette after lunch break endangers one’s babies

      Or, you know, greater exposure of the babies to cigarette smoke?

Leave a Reply