The 4-month sleep regression: what actually changes
Around four months, babies trade newborn sleep for adult-style cycles, and nights get louder before they get better. What the research says, and what helps.
Somewhere between three and five months, a baby who had started sleeping in reassuring stretches begins waking every hour, loudly, as if a switch flipped. Parents call it the 4-month regression. Sleep researchers would call it the opposite, a progression: the moment infant sleep grows up.
What changes at four months?
Newborn sleep has two loosely organized states, and a newborn drops into the deep one almost immediately. Around the fourth month the architecture reorganizes into something much closer to adult sleep: distinct stages, arranged in cycles of roughly 45–60 minutes, each cycle ending in a brief pass through light sleep.
That last part is the whole story. A baby now surfaces into light sleep several times a night by design. Classic infant-sleep research documented decades ago that night wakings do not disappear with age. Babies who “sleep through” still wake between cycles. The difference is what happens next.
Why does it feel like a regression?
Because the wakes are new and the skill is not there yet. Surfacing between cycles, the baby does a quick audit: is anything different from how I fell asleep? A baby who fell asleep feeding, rocking, or held now wakes in a crib that matches none of it, and calls for the missing piece. Every cycle. That is the hourly waking.
Longitudinal work following infants across the first year found that the ability to re-settle without signaling emerges early, for many babies within the first three to four months, when they get opportunities to practice. The regression is loud precisely because the new architecture arrives before the practice does.
What actually helps?
The research consensus is unglamorous: consistency at bedtime, and giving the baby some part in falling asleep. A large multinational study of bedtime routines found a dose-dependent relationship. The more consistent the routine, the better the sleep outcomes, across cultures and ages.
Inside tiho’s two lanes that translates differently, and both work.
- Behavioral. Hold the routine steady and put the baby down where the night will happen, aiming for drowsy-but-awake at least once a day. The schedule is the anchor, and the crib at bedtime should look like the crib at 2 AM.
- Responsive. Keep the cue-response loop steady and predictable: same wind-down, same response to night wakes, so the baby’s audit finds nothing alarming. You are not withholding a response. You are making the response boring.
What does not help: changing everything at once during the loudest week. A regression is the worst possible evaluation window for any method.
What this looks like in the log
In tiho the regression has a signature: total sleep holds roughly steady while wakings spike and settle duration stretches. That distinction matters. The baby is not losing sleep so much as renegotiating how it starts. Watch the weekly trend rather than any single night. When settle times shorten again, the cycles are linking. Most families see the curve bend within two to six weeks.
The night is not broken. It is under construction.
Frequently asked
How long does the 4-month sleep regression last?
The neurological change is permanent, and that is the point. The loud part typically lasts two to six weeks, which is the time it takes a baby to practice linking the new cycles.
Does every baby have a 4-month regression?
Every baby goes through the sleep-architecture change, but not every family notices it. Babies who already settle at bedtime without help often pass through with barely a ripple.
Should we change our sleep method during a regression?
Usually no. A regression is the worst week to evaluate any method, because everything looks broken. Hold your lane for two weeks, keep logging, and judge by the trend, not the worst night.
