At some point around week three of broken sleep, every new parent types the same thing into Google: "when do babies sleep through the night." You're doing it right now, probably. It's 2 AM and you need someone to tell you this won't last forever.

It won't. But the timeline is longer and messier than most people expect, and the phrase "sleeping through the night" means something different to researchers than it does to exhausted parents scrolling their phones in the dark.

What "sleeping through the night" actually means

In clinical sleep research, "sleeping through the night" is defined as five consecutive hours of sleep. Not eight. Not twelve. Five.

That number surprises most parents because it doesn't match the fantasy. When someone says their baby "sleeps through the night," you imagine the child going down at 7 PM and waking up at 6 AM. But by the clinical definition, a baby who sleeps from midnight to 5 AM and then wakes up screaming is sleeping through the night. Congratulations.

This definitional gap matters because it shapes expectations. Studies that report 50-60% of infants "sleeping through" by 6 months are using the five-hour definition. If your standard is an uninterrupted 10-hour stretch, the numbers look very different, and the timeline pushes much later.

For the rest of this article, I'll be clear about which definition is being used. Anytime I say "sleeping through," I mean an uninterrupted stretch long enough that no one has to get out of bed.

Newborns (0-8 weeks): there is no night

Newborns don't have a circadian rhythm yet. Their internal clock hasn't learned the difference between day and night, so they sleep in short bursts of 2-4 hours around the clock, driven by hunger rather than light cycles.

Total sleep in this period is actually high, around 14-17 hours per day, but it's chopped into fragments that don't line up with when you want to be sleeping. A three-hour stretch feels like a gift. A four-hour stretch feels like a miracle.

There's nothing you can do to make a newborn sleep longer stretches, and anyone selling you a product that promises otherwise is not being honest with you. Newborn stomachs are tiny. They need to eat frequently. This is normal and healthy, even though it is brutal.

What you can do during this phase is start introducing day-night cues: keep daytime bright and active, keep nighttime dark and quiet. This doesn't produce immediate results, but it helps the circadian rhythm develop over the coming weeks.

2-3 months: the first longer stretches

Around 6-8 weeks, something shifts. The circadian rhythm starts to emerge. Melatonin production kicks in (newborns don't produce their own melatonin in meaningful amounts). The longest sleep stretch begins drifting toward nighttime.

By 2-3 months, many babies consolidate one longer stretch of 4-6 hours, usually in the first half of the night. They'll still wake to feed after that, but there's now a recognizable pattern: longer sleep early in the night, shorter stretches toward morning.

This is when a simple bedtime routine starts to pay off. The baby's brain is beginning to distinguish "bedtime" from "naptime," and consistent cues (dim lights, a lullaby, a specific sequence of events) help reinforce that distinction.

You're not sleeping through the night yet. But you're sleeping in bigger pieces, and that makes a real difference.

4 months: the regression that isn't really a regression

Just when things start improving, the 4-month sleep regression hits, and it hits hard. Parents describe it as their baby suddenly "forgetting" how to sleep. A child who was doing five-hour stretches is now waking every 90 minutes.

What's actually happening is a permanent change in sleep architecture. Before 4 months, infant sleep cycles are simple: active sleep and quiet sleep. Around 4 months, the brain matures into adult-style sleep cycles with distinct stages (light sleep, deep sleep, REM). Each cycle lasts about 45-60 minutes, and between cycles, the baby passes through a brief period of near-wakefulness.

Adults do this too. We surface between sleep cycles several times a night. The difference is that we've learned to put ourselves back to sleep without fully waking up. A 4-month-old hasn't learned that yet. So at the end of every sleep cycle, they wake up, realize conditions have changed (the rocking stopped, the breast or bottle is gone, the room is darker than when they fell asleep), and they cry.

The 4-month regression isn't really a regression. It's a permanent developmental change. Sleep doesn't go back to the way it was before. The baby needs to learn to connect sleep cycles on their own, and that takes time.

This is the stage where sleep associations become critical. Whatever conditions are present at the moment the baby falls asleep are the conditions they'll expect when they surface between cycles. If they fell asleep nursing, they'll want to nurse again at every wake-up. If they fell asleep hearing a consistent lullaby and then the room went quiet, they'll learn that quiet room = sleep.

5-8 months: when longer sleep becomes possible

For many families, somewhere between 5 and 8 months is when longer stretches start to happen. The baby's stomach is bigger (so they can go longer without eating), and their nervous system is mature enough to connect sleep cycles with some consistency.

This doesn't happen on its own for every baby. Some need a push (what sleep consultants call "teaching independent sleep skills"). Others just gradually start sleeping longer without any intervention. Both are normal.

By 6 months, many pediatricians will tell you the baby is physiologically capable of going 8-10 hours without a feeding. That doesn't mean they will. Capability and habit are different things. A baby who is capable of sleeping 10 hours but has been fed at every wake-up for 6 months has a strong habit to unlearn.

The research here is reassuring but messy. A large longitudinal study published in Pediatrics in 2018 found that 57% of 6-month-olds were not yet sleeping 8 consecutive hours. By 12 months, 28% still weren't. These are healthy, normal babies. Sleeping through is not a milestone with a fixed date. It's a gradual process with a wide range of normal.

8-10 months: separation anxiety enters the chat

Just when sleep starts consolidating, many families hit the 8-10 month regression. This one is driven by a pile-up of developmental milestones: crawling, pulling to stand, separation anxiety, and sometimes the first words. The baby's brain is processing so much new information that sleep gets disrupted.

Separation anxiety is the big one. Around 8-9 months, babies develop object permanence — they understand that when you leave the room, you still exist somewhere else. Before this, out of sight was genuinely out of mind. Now, out of sight means "where did you go and are you coming back?" Bedtime, which involves you leaving, becomes stressful in a new way.

This regression typically lasts 2-4 weeks. The best response is to maintain your existing routine and ride it out. Adding new sleep crutches during a regression (co-sleeping when you didn't before, rocking to sleep when they'd been falling asleep independently) tends to create new habits that outlast the regression itself.

12-18 months: most babies get there

By 12 months, the majority of babies are capable of and often achieving 10-12 hour stretches at night. Not all of them. Not every night. But most, on most nights.

The nap transition around 12-18 months (dropping from two naps to one) can temporarily disrupt nighttime sleep. An overtired baby who missed their second nap but isn't ready for a single-nap schedule might have a rough few weeks at night. This sorts itself out as the new nap schedule stabilizes.

Toddlers who are still waking frequently at night past 12 months often have a habit-based issue rather than a developmental one. They're waking between sleep cycles and don't know how to get back to sleep without intervention. This is very fixable, but it requires changing the sleep associations.

A brief word on sleep training

This article is about timelines, not sleep training methods, so I'll keep this brief. The research on sleep training is actually quite clear: graduated methods (like Ferber/check-and-console) and extinction methods (cry-it-out) both work, both produce results within a week for most families, and long-term studies have found no adverse effects on attachment, behavior, or cortisol levels.

A 2016 randomized controlled trial published in Pediatrics followed families for a year after sleep training and found no differences in parent-child attachment, child emotional regulation, or cortisol levels between the sleep-trained group and the control group. The sleep-trained group just slept better.

Whether to sleep train is a personal decision, and no one should pressure you either way. But if you're suffering and your baby is past 4-6 months, the evidence says it's safe and it works.

What you can actually control

You can't control when your baby's sleep matures. That's neurological development on its own schedule. But you can control the environment and the cues.

A consistent bedtime routine helps the brain anticipate sleep. Same steps, same order, every night. A consistent sleep cue, whether that's a lullaby or a specific phrase, builds the strongest association when it's the last thing before lights out. The sleep environment matters too: dark room, cool temperature (68-72°F), white noise if it helps.

The biggest one, though, is falling asleep independently. That's the skill that lets babies connect sleep cycles without calling for you. Whatever conditions exist when they fall asleep are the conditions they'll expect at 2 AM. If the room was quiet and you weren't there, they'll go back to sleep on their own. If you were rocking them, they'll need you to come back and rock them again.

If you're looking for a sleep cue that stays consistent across every bedtime, that's what Starry Songs does — a short, personalized lullaby with your child's name that sounds the same every night. You play it at the end of the routine, and over time, the song itself starts triggering drowsiness.

The honest timeline

Here's the timeline nobody puts on Instagram:

If your baby is 7 months old and still waking twice a night, that's normal. If your neighbor's baby slept 12 hours at 10 weeks, that's an outlier. Don't measure yourself against outliers.

It does get better. Probably not as fast as you want, and definitely not on a predictable schedule. But it gets better.