Even the "good sleepers" wake every night. I have watched the footage

"Sleeping through the night" does not exist. We have it on camera

Here is my favourite finding in the whole infant sleep literature, because it quietly reverses the thing almost every parent is chasing and almost every professional is asked about. When researchers put infrared cameras on 101 London infants overnight, they found that the babies described as "sleeping through" were not actually sleeping through (St James-Roberts et al., 2015). They were waking in the night like everybody else. The difference was what happened next: they resettled themselves, quietly, without signalling to anyone. The literature now distinguishes "signallers" from "self-resettlers", and on video, around a quarter of babies were already waking and resettling in the night at five weeks old.

So night waking is not the variable that separates "good" sleepers from "bad" ones. Everybody wakes. Waking between sleep cycles is a structural feature of human sleep at every age, including yours. What varies is the behaviour on waking, and that is a completely different clinical question from the one most professionals have been trained to ask.

And the timelines are later than the baby industry suggests. In a cohort of 388 infants, 43% had never, by parental report, managed eight hours of unbroken sleep at six months. At a year old, 43% still had not (Pennestri et al., 2018). The same team then asked a sneakier question: do the "good sleepers" at least do it every night? Across two weeks of nightly sleep diaries, six-month-olds managed eight hours of unbroken sleep on roughly three nights in fourteen on average, and half never did it once in the fortnight (Pennestri et al., 2020). So "sleeping through" is not a milestone a baby reaches and keeps. It is a thing some babies do, on some nights.

That is not a cohort of disordered sleepers. That is the normal distribution, measured.

Unbroken sleep predicts… nothing, as far as we can measure

This is the reversal that genuinely surprises professionals. The pressure on families rests on an assumption so widespread that almost nobody checks it: that consolidated sleep is developmentally necessary, and a baby who wakes is a baby falling behind.

The Pennestri team checked it. They followed those 388 mother–infant pairs and tested whether sleeping through the night (six or eight hour blocks) at six or twelve months was associated with mental development, psychomotor development, or maternal mood, measured at six, twelve and thirty-six months. The association they found was: none. No relationship between unbroken sleep and the Bayley development scales. No relationship with maternal mood either (Pennestri et al., 2018). The one robust association in the data pointed the other way entirely: the babies who woke more were more likely to be breastfed, which is exactly what feeding biology would predict.

A five-year randomised follow-up tells the same story from the opposite direction: children whose parents used behavioural sleep interventions in infancy were indistinguishable from controls at age six, on harms and on benefits (Price et al., 2012). The honest headline from both directions is that night waking in infancy is not the developmental emergency it is marketed as.

Which changes what we are actually treating when a family asks for help: usually parental exhaustion, expectation and capacity. Real problems. Different problems.

The famous cortisol study does not show what either side says it shows

If you have spent any time in this field online, you have met the cortisol argument: babies left to cry stop signalling but remain physiologically stressed, "proven by science." The study behind it is real and worth knowing properly. Middlemiss and colleagues (2012) sampled salivary cortisol in 25 mother–infant pairs during a five-day inpatient extinction programme. By the third day, the infants had stopped crying at the sleep transition, but their cortisol remained elevated while their mothers' levels fell: mother and baby had become physiologically out of sync.

That is a genuinely important observation about behavioural quietness not equalling physiological calm. But here is what the study cannot tell you, because of how it was built: there was no control group, no home baseline, and the setting was an unfamiliar residential unit, which raises cortisol on its own. Twenty-five babies, two time points, no comparison condition. It cannot establish that extinction causes harm.

And before the other camp celebrates: the trials run in its favour are nearly as thin. The most-cited randomised trial of graduated extinction found no elevated cortisol and no attachment differences at follow-up, in a total sample of 43 infants (Gradisar et al., 2016). A longitudinal study found no adverse effects of parent-reported "cry it out" on attachment or behaviour at 18 months (Bilgin & Wolke, 2020), observationally. The honest reversal is this: the evidence base both camps shout about is a handful of small, short studies.

An evidence-based practitioner is not someone who can cite one of them. It is someone who knows the width of the error bars, and practises accordingly.

Self-settling is not taught. It is developed, and relationships build it

"You have to teach them to self-settle" may be the most repeated sentence in infant sleep, and the video evidence quietly reverses it. In the St James-Roberts cohort, self-resettling was already present in a quarter of five-week-olds, long before any teaching is possible. The babies who resettled at five weeks were the ones most likely to be sleeping five-hour stretches at three months. And the breastfed babies were just as likely to self-resettle as the formula-fed ones (St James-Roberts et al., 2015), which deletes another piece of received wisdom while it is at it. The capacity looks developmental: it emerges, on its own timetable, the way rolling and babbling do.

What does the adult contribute, if not teaching? The best answer in the literature is relational. Maternal emotional availability at bedtime, observed on video, predicts infant sleep quality (Teti et al., 2010). Not a technique. Not a settling method. The emotional quality of what happens around sleep. Which is why, in my framework, we spend far more time on the relationship and the context around sleep than on the settling strategy, and why the strategy is almost the last thing we choose.

Self-settling is not taught. It emerges. And relationships are what it emerges from.

The averages cannot tell you about the baby in front of you

One more reversal, and it is the one that makes the others usable. Everything above is population data, and population data has a limit that sleep charts quietly ignore: the spread. A systematic review of normal infant and child sleep found variability so wide that the authors cautioned against using averages as individual targets at all (Galland et al., 2012). The Zurich longitudinal percentiles put the normal range for total sleep at six months at roughly ten to eighteen hours a day (Iglowstein et al., 2003). An eight-hour spread, within normal. And culture moves the goalposts again: across 17 countries, average toddler bedtimes differ by around two hours, with no corresponding epidemic of dysfunction (Mindell et al., 2010).

So when a chart says "babies this age need 14 hours", the science says: some babies this age need 14 hours. The only way to find out what this particular child needs is a skilled individual assessment: sleep pressure, temperament, feeding, health, development, family context.

That is a clinical skill. It can be learned. And it is the entire reason holistic sleep coaching exists.