Skip to content
Kura

12 things with strong or good evidence

What the evidence supports

Twelve things with several trials, or a large and consistent body of evidence, behind them. Most of them are unglamorous. Each card opens up into the mechanism, the key studies (with their numbers), what it looks like at 3 a.m., and the caveats.

  1. Strong Responsive caregiving

    Notice what the baby does, and answer it.

    When she looks, coos, reaches or cries, respond within a few seconds with your face, voice or touch. That loop is the core skill, and most of the rest of this site turns out to be some version of it.

    Researchers call this back-and-forth serve and return (the baby serves, you return), and done consistently it produces what they call secure attachment: the baby’s working assumption that someone reliable is there. Coaching programmes that teach exactly this have been tested in randomised trials, and they produce more securely attached babies and healthier stress-hormone patterns. The effects are moderate rather than miraculous. Interestingly, when the studies were pooled, short and focused coaching beat the long intensive programmes.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    A baby’s brain builds its expectations of the world out of thousands of tiny exchanges a day. Each time a cue gets a timely answer, she learns that signalling works and that arousal comes back down, which is the beginning of emotional regulation (and, later on, the same loop shows up as language, curiosity and self-control).

    You can see the expectation in the lab. In the “still face” experiment a parent plays normally and then holds a blank face for two minutes; three-month-olds first work quite hard to get the parent back, then turn away and fall apart. Put differently, babies are not passive. They are already running the loop.

    The evidence

    The pooled analysis of 70 coaching studies found that programmes teaching parents to read and answer cues improved parental sensitivity by a moderate amount and attachment security by a small one. Programmes with fewer than 16 sessions did better than longer ones, which the authors titled “less is more“.

    Attachment and Biobehavioral Catch-up (ABC), ten home visits with a coach who comments in the moment (“she looked at you and you smiled back“), was tested in a randomised trial with 120 high-risk families. Children of coached parents were far more likely to be securely attached, and a second trial found that their daily stress-hormone rhythm had normalised.

    The strongest evidence of all comes from the absence of responsiveness. In the Bucharest Early Intervention Project, institutionalised infants randomised to foster families gained about eight IQ points by age four and a half, with the largest gains for those placed before 24 months. Neglect is the most damaging thing in the whole literature, which is another way of saying that ordinary responsiveness is the most protective.

    Caveats

    “Responsive” does not mean instant or constant. Waiting a few seconds, and letting an older baby work on a small frustration, is part of the loop.

    Attachment security is one of several outcomes, and its long-term effects are modest once you pool the studies. It is a foundation rather than a guarantee.

  2. Strong Language · Early learning

    Talk with the baby, and leave room for her to answer.

    Back-and-forth exchanges, even before words, are what build language. Live faces count; recordings, as far as anyone can tell, do not.

    The famous “30 million word gap” was based on 42 families and has not held up under re-analysis. What did hold up is that the number of back-and-forth exchanges (conversational turns) predicts language skill, and even activity in the brain’s language areas, independent of family income. The mechanism appears to be live interaction: nine-month-olds who spent twelve sessions with a live Mandarin speaker learned its speech sounds, while babies given the identical sessions on video or audio learned nothing at all.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Language is learned socially. A baby is tracking where you look, what you respond to, and the rhythm of the exchange, and a turn (she vocalises, you answer within a few seconds) gives her timing and feedback that a stream of words does not. This is why the “narrate everything” advice is about half right: narration helps most when something the baby just did prompted it.

    The evidence

    In the MIT study, 36 children aged four to six wore recorders for two days. Conversational turns, rather than total words heard, predicted verbal scores and activation in Broca’s area during a language task, and family income did not explain the effect.

    The Kuhl experiment is the clearest demonstration I have seen of the live-interaction requirement. Twelve 25-minute sessions of Mandarin from a live tutor let American nine-month-olds hear Mandarin sound contrasts as well as Taiwanese babies did; the video and audio groups were indistinguishable from babies with no exposure.

    The original word-gap study recorded only parent-to-child speech in 42 families. A 2019 re-analysis that recorded whole households found no consistent gap by income. The lesson, as I read it, is that quality and reciprocity matter more than volume - not that talk does not matter.

    Vocabulary at three predicts reading in third grade, so the payoff is a long one.

    Caveats

    Most of this evidence is observational (families who talk more differ in other ways too). The Kuhl and dialogic-reading trials are the parts with experimental proof of mechanism.

    The point is turns rather than performance. A quiet parent who answers cues does more than a chatty one who talks over them.

  3. Strong Language · Early learning

    Read aloud from the start, and let her interrupt.

    A few minutes a day with a book, asking rather than reciting, has trial evidence behind it.

    Programmes that hand families a book and two minutes of coaching at each check-up (the Reach Out and Read model) improved children’s language in randomised and controlled studies, and reading that asks questions and follows the child’s answers (dialogic reading) beat straight reading in a trial. For a newborn, the book is mostly a reason to hold her close and talk in a rhythm. The questions come later.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Books do three things a conversation about the room does not: (1) they introduce words that never come up in daily life (“hippopotamus“, “lantern“); (2) they create a shared point of attention that the both of you can point at; and (3) they build a ritual that survives exhaustion. I suspect the ritual is why it works - it is talk that happens even on the bad days.

    The evidence

    In the randomised clinic trial (205 families), parents given books and brief guidance at well-child visits read more, and their toddlers had larger vocabularies. A controlled study of 122 inner-city families found higher receptive and expressive language in children exposed to the programme.

    The dialogic reading trial taught parents to ask open questions, follow the child’s answers, and expand on them. After a month, children in the trained group were six to eight months ahead in expressive language.

    Caveats

    The evidence is for language rather than for reading earlier. Early-reading programmes for infants have no support and some harm (see the avoid list).

  4. Strong Health · Good health

    Hold her skin to skin, early and often.

    Bare chest to bare chest, especially in the first hours and for small babies. One of the strongest findings in the field.

    For small or premature babies, continuous skin-to-skin holding (kangaroo mother care) is one of the strongest findings in all of newborn medicine: a WHO trial across five countries found that starting it immediately, rather than after the baby had been stabilised, cut deaths in the first month by about a quarter in babies under 1.8 kg. For healthy full-term babies the benefits are smaller but real (a better breastfeeding start, steadier temperature and blood sugar, less crying), and they hold whether or not you are breastfeeding.

    Of 100 babies born weighing 1.0–1.8 kg, how many died in the first 28 days

    Skin-to-skin after stabilisation

    16 in 100 (15.7%)

    Skin-to-skin immediately

    12 in 100 (12.0%)

    Each dot is one baby. Filled dots died. The trial was stopped early because the difference was clear. Source: WHO Immediate KMC Study Group (2021).

    View as a table
    Groupbabies who died
    Skin-to-skin after stabilisation16 in 100 (15.7%)
    Skin-to-skin immediately12 in 100 (12.0%)
    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    A newborn’s systems for holding temperature, breathing rhythm, heart rate and blood sugar are all steadier against a parent’s chest than in a cot or an incubator; the parent’s body is, in effect, the thermostat. Skin contact also triggers the hormones that start milk and calm the both of you, and for a preterm baby it reduces exposure to hospital infections.

    The evidence

    The WHO immediate kangaroo mother care trial randomised 3,211 babies weighing 1.0–1.799 kg across Ghana, India, Malawi, Nigeria and Tanzania. Deaths by day 28 were 12.0% with immediate skin-to-skin versus 15.7% with the standard practice of waiting until the baby was stable, and the trial was stopped early for benefit.

    The Cochrane review of kangaroo care for low-birth-weight babies (21 trials) found lower mortality, less severe infection and hypothermia, and better weight gain and breastfeeding.

    For healthy full-term babies, the Cochrane review of 46 trials found that skin-to-skin right after birth increased breastfeeding at one to four months and improved blood-sugar and heart-lung stability. The evidence quality was rated low to moderate, mostly because the trials were small.

    Caveats

    The dramatic mortality effect is in small, mostly preterm babies in lower-resource hospitals. For a healthy full-term baby the gains are real but modest.

    A sleepy parent on a sofa with a baby on their chest is a fall and suffocation risk. Do it awake.

  5. Strong Safety · Safety and security

    On her back, on a flat empty surface, in your room.

    Back to sleep, her own firm flat surface, nothing soft, and the same room as the parents for at least six months.

    Sudden infant death syndrome (SIDS) fell by more than half in the years after the 1994 Back to Sleep campaign. The current American Academy of Pediatrics (AAP) guidance adds a firm, flat, non-inclined surface; no pillows, bumpers, blankets or soft toys; room-sharing without bed-sharing for at least six months; no smoke exposure; and a pacifier offered at sleep. This is survival rather than development, but nothing else on this site matters much if this part goes wrong.

    130
    SIDS deaths per 100,000 births, 1990
    38
    per 100,000 births, 2020
    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Most sleep deaths happen when a baby’s face ends up against something that traps exhaled air, or when a baby who cannot yet lift or turn her head is placed face down. Back-sleeping, a bare firm surface and a cool room remove those mechanisms. Room-sharing appears to work partly because parents hear the baby and partly because babies rouse more easily near their parents.

    The evidence

    US SIDS rates fell from about 130 per 100,000 live births in 1990 to under 40 by the late 2010s, with the steepest drop following the 1994 Back to Sleep campaign. Deaths from all sudden unexpected causes have plateaued since around 2000, which is why the guidance now emphasises the whole sleep environment and not just position.

    A study of about 20 million US births found that any maternal smoking in pregnancy roughly doubled the risk of sudden unexpected infant death, and the risk rose per cigarette.

    The 2022 AAP policy statement reviews the full evidence base, and explains why inclined sleepers, weighted sleep sacks and bedside “in-bed” sleepers are not recommended.

    Caveats

    Rates vary a great deal by country, and the guidance here is the US one. NHS and Canadian guidance agree on the essentials.

    Bed-sharing is common worldwide, and the risk is concentrated where there is smoking, alcohol, sofas, soft bedding, or a very young or preterm baby. The AAP still advises against it; if a family chooses it anyway, the policy statement describes how to make it less dangerous.

  6. Strong Nutrition · Adequate nutrition

    Breastfeed if you can, and do not lose sleep over the IQ claims.

    Clear health benefits. The cognitive benefit is small, and it had faded by adolescence in the one big trial. Formula-fed babies are not behind.

    The health case is clear: fewer infections, a lower SIDS risk, and protection against a dangerous gut disease in preterm babies (necrotising enterocolitis, or NEC). The cognitive case is weak. The one large randomised trial, PROBIT in Belarus, found about six IQ points at age six with a wide margin of error, teachers noticed no difference, and by sixteen the gap had mostly gone. In short - breastfeeding is a good thing for health, it is not what makes a child clever, and a formula-fed baby is not at a developmental disadvantage.

    +5.9
    IQ points at age 6.5 in the PROBIT trial (wide uncertainty)
    ≈0
    meaningful difference at age 16
    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Breast milk carries antibodies and living cells tuned to the mother’s environment, which is the mechanism behind the infection findings. The claimed cognitive benefit was always confounded: in most countries the mothers who breastfeed longer are also older, more educated and wealthier, and only a trial that randomises can separate the milk from the mother. Only one large trial has.

    The evidence

    PROBIT randomised 31 maternity hospitals in Belarus to a breastfeeding-promotion programme, which raised rates of exclusive breastfeeding. At 6.5 years the promotion group scored about 5.9 IQ points higher, but the confidence interval ran from roughly minus one to plus thirteen, and teachers’ academic ratings did not differ.

    At 16, the same children showed no meaningful differences on most cognitive tests, with a small remaining edge in verbal function only.

    The health benefits, by contrast, are consistent across trials and cohorts, and they are why every major body recommends breastfeeding where possible.

    Caveats

    The IQ claim keeps circulating because large observational studies find associations. Sibling comparisons and the randomised trial suggest most of that is the mother rather than the milk.

    Skin-to-skin (the previous card) is the part of “breastfeeding” with the clearest short-term evidence for the baby, and it works with a bottle too.

  7. Strong Nutrition · Adequate nutrition

    Vitamin D, iron from six months, allergens early.

    Three specific and admittedly boring actions with strong evidence behind them, and easy to miss in the fog.

    Three things with strong evidence and low effort: (1) vitamin D drops for breastfed babies, 400 IU a day, from the first days; (2) iron-rich foods from around six months, because iron deficiency in infancy has lasting effects on attention and learning; and (3) early introduction of allergens - the LEAP trial cut peanut allergy by about 80% in high-risk babies when peanut was introduced between four and eleven months, reversing decades of avoidance advice.

    Of 100 high-risk babies, how many had a peanut allergy at age 5

    Avoided peanut

    17 in 100 (17.2%)

    Ate peanut from 4–11 months

    3 in 100 (3.2%)

    Each dot is one baby with severe eczema or egg allergy. Filled dots developed peanut allergy. An 81% relative reduction. Source: Du et al. (2015).

    View as a table
    Groupbabies with peanut allergy
    Avoided peanut17 in 100 (17.2%)
    Ate peanut from 4–11 months3 in 100 (3.2%)
    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Breast milk is low in vitamin D and babies are kept out of the sun, so rickets reappeared when supplementation lapsed. Iron stores from birth run out around six months, just as the brain is laying down myelin and dopamine systems that need it. For allergens, the immune system appears to learn tolerance through the gut during a window in the first year, and avoidance, it turns out, was teaching it the wrong lesson.

    The evidence

    The AAP’s vitamin D guideline set 400 IU a day for all breastfed and partially breastfed infants, beginning in the first few days of life.

    The AAP iron guideline recommends iron-rich complementary foods from about six months, iron drops for exclusively breastfed babies from four months until then, and a blood check at twelve months. The review by Lozoff and colleagues documents lasting attention and motor differences after infant iron deficiency even when it is later corrected.

    LEAP randomised 640 babies aged 4–11 months with severe eczema or egg allergy to eat at least six grams of peanut protein a week or to avoid peanut until age five. Peanut allergy developed in 17.2% of avoiders and 3.2% of eaters, and the result produced new guidelines within two years.

    Caveats

    LEAP studied high-risk babies. Guidelines extend the advice to all babies, with a clinician check first for the high-risk group.

    Iron drops taste bad and can stain teeth, so the food route is preferred once solids start.

  8. Good Movement · Good health

    Tummy time, in short bursts, while she is awake and watched.

    Modest but consistent evidence for motor development and fewer flat spots. Aim for about 30 minutes a day, spread out.

    Because babies now sleep on their backs, they need awake time on their fronts to build neck, shoulder and core strength. A systematic review of sixteen studies found tummy time associated with better gross-motor development and fewer flattened heads (positional plagiocephaly), though the authors rated the evidence low quality. The WHO suggests at least 30 minutes a day, spread across the day, for babies who are not yet mobile.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Lifting the head against gravity is how a baby develops the neck and trunk control that rolling, sitting and crawling depend on. Time on the front also takes pressure off the back of the skull, which is soft and reshapes easily in the first months.

    The evidence

    The 2020 systematic review pooled 16 studies with 4,237 infants. Tummy time was positively associated with gross-motor and total development, the ability to move while prone, and prevention of skull flattening, with low-quality evidence overall.

    WHO’s 2019 under-five guidelines recommend interactive floor-based play several times a day and at least 30 minutes of tummy time spread through the day for infants who are not yet mobile.

    Caveats

    The evidence is low quality and the effect sizes modest. Babies who hate tummy time and get carried upright a lot still develop normally.

  9. Strong Responsive caregiving

    Treating a parent’s depression is also a child intervention.

    Postpartum depression in either parent is linked to poorer outcomes for the baby. Screening and treatment work.

    About one mother in eight and one father in ten develops depression around a birth. A systematic review of 122 studies links maternal postpartum depression to poorer bonding, more infant sleep problems, and poorer language, cognitive and behavioural outcomes, and depression in fathers at eight weeks predicted behaviour problems at age three and a half in a study of 13,000 families. Treatment works, which is why the AAP tells pediatricians to screen mothers at the one-, two-, four- and six-month visits. For what it’s worth, this is the card on the site I would least want anyone to skip.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Depression flattens exactly the thing on the first card: the ability to notice and answer a baby’s cues. It also disturbs sleep, feeding and the parent’s own recovery. The link to child outcomes has nothing to do with being a bad parent; it is an illness that interrupts responsiveness, and one that lifts when it is treated.

    The evidence

    The 2019 systematic review found consistent associations between maternal postpartum depression and worse infant outcomes across bonding, breastfeeding, sleep, and later cognitive, language and behavioural development.

    In the Avon Longitudinal Study, paternal depression at eight weeks roughly doubled the odds of conduct problems in the child at age three and a half, independent of the mother’s depression.

    The AAP’s 2019 clinical report recommends routine screening of mothers at the 1, 2, 4 and 6-month well-child visits and describes referral pathways. Cognitive behavioural therapy, interpersonal therapy, peer support and medication all have trial evidence behind them.

    Caveats

    These are associations from observational studies, and parents with depression face other stresses too. That said, treatment trials show that treating the parent improves parent-child interaction, which is the pathway.

  10. Strong Health · Good health

    Keep the visits, the hearing test and the milestone checks.

    This is where timing genuinely matters. Hearing loss and autism found early have very different trajectories from those found late.

    Most of the “critical window” anxiety in parenting is overblown, and this card is the exception. Babies whose hearing loss is identified and treated before six months develop language close to their hearing peers; identified later, they usually do not. Toddlers with autism who started intensive play-based intervention early gained about eighteen IQ points in a randomised trial, versus seven for the comparison group. The well-child schedule, with its milestone questions and screens, exists to catch these.

    1 · 3 · 6
    months: hearing screened, diagnosed, intervention started
    +18 vs +7
    IQ points over two years, early autism intervention vs usual care
    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Language, vision and social attention are built on input that arrives during specific early periods. A baby who cannot hear speech in the first year misses the window in which the brain tunes itself to the sounds of her language, and catching up later is much harder. Autism intervention works partly because it teaches parents to create serve-and-return loops with a child who does not initiate them.

    The evidence

    Yoshinaga-Itano’s study of children with hearing loss found that those identified before six months had language quotients close to the normal range at every age tested, regardless of the severity of loss, family income or other factors. Those identified later did not.

    The Joint Committee on Infant Hearing sets the 1-3-6 benchmarks: screened by one month, diagnosed by three, intervention by six.

    The Early Start Denver Model trial randomised 48 toddlers aged 18–30 months with autism to two years of intensive play-based intervention or community care. IQ rose about 18 points versus 7, with better adaptive behaviour.

    The AAP recommends developmental surveillance at every visit, standardised screening at 9, 18 and 30 months, and autism-specific screening at 18 and 24 months. Vaccines on schedule belong in this card too, since the diseases they prevent (measles and Hib meningitis among them) damage developing brains.

    Caveats

    Milestone ranges are wide, and a late walker or talker is usually just a late walker or talker. The point is to ask rather than to panic.

  11. Strong Responsive caregiving

    Do not hit. Warm and firm carries the day; harsh does not.

    Spanking is associated with worse outcomes on 13 of 17 measures and better outcomes on none. Limits without pain work better.

    A pooled analysis of 75 studies covering about 161,000 children found spanking associated with worse outcomes on 13 of 17 measures (aggression, mental health and the parent-child relationship among them) and with better outcomes on none, and the AAP formally opposes spanking and harsh verbal punishment. The style that does work, in infancy and for the eighteen years after it, is warm plus firm: clear limits, explained, and enforced without fear.

    Of 17 child outcomes studied, spanking was associated with
    Worse
    13 worse outcomes
    No clear link
    4 no clear link
    Better
    0 better outcomes

    Pooled from 75 studies. Effect sizes were small to moderate, and all pointed the same way. Source: Gershoff et al. (2016).

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    Hitting teaches that the bigger person’s frustration ends the conversation. It raises the child’s stress response in the moment and, over time, models aggression as a tool. Meanwhile, it does not even do the job: compliance after spanking is no better than after non-physical limits, and it erodes the relationship that makes limits work in the first place.

    The evidence

    Gershoff and Grogan-Kaylor’s 2016 meta-analyses separated spanking from more severe physical abuse and still found the same direction of effect across 13 of 17 outcomes, in both cross-sectional and longitudinal studies.

    The AAP’s 2018 policy statement reviews the evidence and recommends positive discipline strategies: praise for good behaviour, limit-setting, redirection, and time-outs used calmly for older toddlers.

    Longer term, Steinberg’s review of decades of adolescent research finds warm-plus-firm (“authoritative“) parenting consistently linked with the best outcomes across cultures.

    Caveats

    The studies are observational, and critics argue that children who are already difficult get spanked more. Longitudinal studies that control for earlier behaviour still find the effect, though the size is modest.

  12. Strong Safety · Safety and security

    Keep smoke, alcohol and lead away from her.

    Three exposures with large, well-documented effects on the developing brain and body. Removing harm does more than adding enrichment.

    The biggest effects in the whole literature come from removing harm rather than from adding enrichment, and these three are the clearest cases. Smoking around a baby roughly doubles the risk of sudden unexpected death. Alcohol in pregnancy is the leading preventable cause of intellectual disability. And lead (mainly from paint in homes built before 1978, and from old pipes) lowers IQ most steeply at the lowest exposures, which is why there is no safe level.

    Read on: why it works, the evidence, the caveatsShow less

    Why it works

    All three act on a brain that is growing faster than it ever will again. Lead substitutes for calcium in developing neurons, alcohol disrupts cell migration in the fetal brain, and smoke exposure impairs a baby’s arousal from sleep while damaging the airways. None of them can be offset by anything on the “what works” list.

    The evidence

    The US birth-cohort study of about 20 million births found any maternal smoking in pregnancy associated with roughly double the risk of sudden unexpected infant death, rising with cigarettes per day. Second-hand smoke after birth is an independent risk factor in the AAP’s safe-sleep review.

    The pooled analysis of 1,333 children across seven studies found that IQ losses per unit of blood lead were largest at the lowest levels. The CDC lowered its reference value to 3.5 micrograms per decilitre in 2021 and states that there is no known safe level.

    CDC and every major body agree that no amount of alcohol in pregnancy has been shown to be safe, and fetal alcohol spectrum disorders remain the leading preventable cause of intellectual disability.

    Caveats

    Lead and smoke risks depend heavily on housing and geography. In a new build with no smokers, this card is done.

Next: the claims you will hear that the evidence does not quite support.

Mixed and oversold