somewhere around eight weeks, some babies do something remarkable. they give you a five hour stretch. sometimes six. you go to bed braced for the usual and you wake up genuinely confused about what time it is.
and then around ten weeks it disappears.
i get a version of this message almost every week.
you didn’t do anything wrong. the stretch you lost was never a skill your baby had. it was a side effect of not having a body clock yet.
one thing before the explanation, because i’d rather you know where it comes from. the eight-to-ten-week shape is a pattern i see in my inbox. it isn’t something my training covers, and an inbox is not a representative sample. the research says only about eight percent of babies have even one five hour night by two months. so this isn’t most babies. it’s the ones whose parents end up writing to someone like me.
what’s underneath it is well documented, though. that part i can explain properly.
what was actually happening
newborns don’t make their own melatonin. they borrow yours through the placenta before birth, and they don’t start producing it themselves until somewhere around three to four months. the circadian rhythm, the thing that tells a body it is night, doesn’t really come online until then, and it isn’t fully mature until five or six months.
so at eight weeks, your baby was not sleeping through part of the night because she had learned that night is for sleeping. she was sleeping because she hit a long run of deep sleep early in the night, and deep sleep is where almost everything that looks like sleeping through the night actually happens. one night out of a very variable spread of nights. not a pattern. a number at the good end of a range that was always that wide.
around ten weeks a few things arrive at once. alertness climbs. and her cycles are still short, about fifty minutes, with a surfacing point at the end of each one. that part doesn’t change at ten weeks and it doesn’t change at four months either. what changes is whether she sleeps through those points by accident or notices them.
the bigger reorganization, the one where newborn sleep starts becoming the sleep she’ll have for the rest of her life, is closer to three or four months than to ten weeks. what’s happening now is the run-up to it.
why it feels like failing
because you had evidence. that’s the cruel part. if she had never given you five hours you would have no expectation to be disappointed by. instead you got a preview, you built a story around it, and then the story broke.
the story usually goes: she could do it, so she should be able to do it, so something is stopping her, so it’s probably me.
nothing is stopping her. she is ten weeks old.
one caveat, and i mean it as the useful kind. everything in this letter assumes a baby who is feeding well and growing well. if the nights got worse and stayed worse, or if she seems uncomfortable rather than just awake, that is a call to your pediatrician before it is a change to your rhythm. a routine can’t fix discomfort. trust the people who tell you where their expertise ends.
what to do about it
not sleep training. that’s a four month and up conversation, and for good reason, because the thing you would be training doesn’t exist yet.
what actually helps between now and then:
light in the morning. outside, or at least a bright window, early in that first wake window. light is what sets the clock. what you do now doesn’t speed up the calendar. it builds the conditions the clock emerges into. it feels too simple to matter. it matters.
cap the naps. if a daytime nap runs past two hours, wake her, assuming your pediatrician hasn’t asked you to do something different with her feeds. you’re not stealing sleep, you’re moving it. a long daytime nap spends the sleep pressure she needs for the night.
count the wake window from eyes open to eyes closed, not from eyes open to when you start trying to get her down. the wind-down is inside the window, not added on after it, and at ten weeks the window is short enough that the difference is most of it. this is the most common counting error i see.
keep the evening boring. same order every night, lights low, feed early in the routine rather than last. not because feeding to sleep is a problem at ten weeks, it isn’t. because the order is the signal.
let her practice going down awake sometimes. not every time, not as a project. occasionally, put her down drowsy and awake and see what happens. expect practice, not performance.
the honest timeline
between four and six months the circadian rhythm finishes coming online, the startle reflex fades, sleep cycles mature, and she becomes genuinely capable of learning to fall asleep on her own. that’s when the work becomes possible, and when it becomes fair.
until then you are not behind. you’re in the part where the machinery is still being built, and there’s no way to hurry it. only ways to feed it.
the long stretch comes back, and when it does it comes back differently. it’ll be something she can do rather than something that happened to her. that’s the version worth waiting for.
you’re not back at the start. you never had the thing you think you lost.
so: how old was she when the stretch went, and what have you been telling yourself about it since?
these letters are free. one at a time, the reasoning under the advice, written to be read in a nap window. subscribe here.
working 1:1 starts with a 10-minute intake, and i read every one before we ever talk. tell me about your child.
✶ and for the record, you’re doing great.

