If your baby has just turned four months old and is suddenly waking every ninety minutes when she used to do four-hour stretches, you are not doing anything wrong. The four-month sleep regression is one of the most predictable developmental events in the first year, and it is the consequence of a permanent rewiring of your baby's sleep architecture. It is not a phase she will grow out of and return to her old patterns — the new patterns are the adult-like ones, and once they arrive they stay.
This is a guide written for the woman sitting on the edge of the bed at 3 a.m., second-guessing every decision she has made in the last four months. The biology, the strategies that have evidence behind them, and the advice that, as far as the research goes, makes very little difference one way or the other.
What is actually happening in there
Newborns sleep in two simple states — active and quiet — and cycle through them roughly every fifty minutes. Around the third and fourth month, that simple system is replaced by adult sleep architecture: four distinct stages including light sleep, deep sleep, and REM. The transitions between stages create natural wake points roughly every forty-five to sixty minutes. An adult barely notices these. A baby who has just acquired the new architecture wakes fully at every transition because she has not yet learnt to fall back asleep on her own.
That is the entire mechanism. The baby is not in pain, not hungry in a new way, not regressing in any developmental sense. Her brain has matured, and the maturation has temporarily made her a worse sleeper. The window of disrupted sleep usually lasts two to six weeks while she learns the new skill of resettling.
The timing is not exactly four months. About half of babies hit it between fifteen and eighteen weeks. A quarter hit it earlier. A quarter hit it later, sometimes as late as twenty weeks. If your baby is six months old and still sleeping well, you may have slipped through. It is rare but not impossible.
The strategies that have evidence
The research on infant sleep is messier than the research on adult sleep — small samples, variable definitions, lots of self-report — but a few interventions have replicated reasonably well.
A consistent bedtime routine. The same sequence, in the same order, in roughly the same time window. Bath, feed, book, song, dark room. The evidence here is strong: babies whose parents follow a routine of even three or four steps fall asleep faster and resettle faster between sleep cycles by six to eight weeks of consistent practice. The specific steps matter less than the consistency.
A dark room. Blackout curtains. Genuinely dark, not curtain-with-some-light dark. Melatonin production in the baby's brain at this age is exquisitely sensitive to light, and even faint ambient light from a hallway or a streetlamp can delay her ability to consolidate sleep. Pound for pound, blackout curtains are one of the highest-leverage purchases of the first year.
White noise. A continuous low-frequency hum at roughly fifty to sixty decibels, similar to a vacuum cleaner running in the next room. White noise masks the small environmental sounds that wake babies at the transitions between sleep cycles. A dedicated machine — the Hatch Rest or the Yogasleep Dohm are the two most commonly recommended — beats a phone app, mainly because phones go to sleep and machines do not.
Gentle sleep training, started after week one of the regression. The catch-all phrase covers several methods, from the gradual ones where you sit by the cot and slowly move further away each night, to the more abrupt ones such as Ferber's check-and-console. The published comparison studies are remarkably consistent: nearly all forms of structured sleep training reduce night wakings by week three, and none of them, by the time children are followed up at five and ten years, produce measurable differences in attachment, behaviour, or mental health. This is one of the most replicated findings in infant research and yet the most denied one online.
If you decide to sleep train, the four-month regression is the natural window. Earlier than fourteen weeks is too soon — the sleep architecture is not yet mature. Later than seven months is harder because the baby's protest mechanisms have grown more sophisticated.
What does not help, despite being recommended everywhere
Adding rice cereal to the last bottle. The trial evidence on this is decades old and consistently negative. Babies who get rice cereal at four months do not sleep longer; they do, on average, take in more calories per day and gain weight faster, which has its own implications. It is a piece of generational advice that has not held up to scrutiny.
Lengthening the wake windows aggressively. The wake-window obsession on social media has more to do with monetised sleep coaches than with infant biology. Some babies do well on ninety-minute wake windows at four months; some need two hours; some can stretch to two and a half. Watch your baby, not a chart.
Buying a smart-cradle product. The expensive cradles that rock and respond to the baby have very mixed evidence. Some parents swear by them, particularly with reflux babies. Many find them no more useful than a £30 baby monitor and a quiet room. At £1,200 a unit, the cost-per-night-of-sleep is high enough that you should be cautious about expectations.
The cry-it-out wars on Instagram. The methods are real, the controversy is largely manufactured, and your baby will not remember this period. Pick a strategy you can stick to consistently for two weeks. Inconsistency is what fails, not the choice between methods.
How to survive the parental side of it
The regression is harder on parents than on babies. Babies recover from interrupted sleep within minutes; adults need consolidated cycles to feel functional, and four-month-regression parents go three to six weeks without one.
Split the night with your partner if you have one. Even non-breastfeeding partners can take the first stretch of the evening — say 7 p.m. to 1 a.m. — while you sleep through it. You take the second half. Single mothers do not have that option and need to be aggressive about asking for help — a grandparent for a weekend, a friend for an afternoon. Sleep debt at this stage is a serious risk factor for postnatal depression and is taken seriously by your GP if you mention it.
Cancel almost everything for the duration. The regression is six weeks at most. Whatever social or work plans you cancel will be there afterwards. The marathon training, the renovation project, the side hustle — all of them can wait. The regression is short relative to a year, but it is long relative to your tolerance for sleep deprivation, and protecting your nights buys you everything else.
If you are not coping — properly not coping, not just tired — tell your health visitor or GP. Postnatal depression presents most often at three to six months, not in the early weeks, and the four-month regression is the trigger for many mothers. There is no medal for getting through it alone, and the treatment is good if you ask for it.