Sleep regressions by age, and how long each one lasts

“Regression” is a misleading word. Nothing is going backwards — the sleep is disrupted precisely because something is moving forward. A baby learning to roll, crawl, stand or talk will often practise it at 2am, and their brain is busier at night while the skill consolidates.

That framing matters, because it changes what you do about it. You are not fixing a fault. You are keeping things steady while a phase passes.

Sleep regression ages at a glance

Each regression, age by age

The ages above are averages of when the underlying skill usually arrives, not appointments. An early crawler has their crawling regression early. What follows is what each one usually looks like from the inside.

The 4-month sleep regression

The only one that is not a phase. Sleep architecture matures into cycles, and a baby who used to sleep through a nap transfer now surfaces every 45 minutes and needs help getting back down. Expect frequent night waking, naps that shorten to a single cycle, and a bedtime that suddenly takes twice as long. Two to six weeks of adjustment; the change itself is permanent.

The 6-month sleep regression

Less universal than the 4-month one. It usually tracks rolling and sitting — a baby who can now roll will practise it in the cot at 2am — and it sometimes lands on top of starting solids. Early rising is the common complaint. Two to four weeks.

The 8-month sleep regression

Often the longest, because three things arrive together: crawling or pulling to stand, separation awareness, and frequently a nap that is ready to be dropped. Standing up in the cot and being unable to get back down is the signature. Three to six weeks, and the one most likely to be mistaken for a habit problem.

The 12-month sleep regression

Walking, first words, and often the two-to-one nap transition circling. Nap refusal and bedtime resistance dominate; nights are usually less affected than at eight months. Two to four weeks. Resist dropping the second nap on the evidence of a bad fortnight.

The 18-month sleep regression

A language leap, molars coming through, and a toddler who has discovered that bedtime is negotiable. Bedtime battles and night waking with genuine protest rather than distress. Two to six weeks, and the one where consistency matters most, because habits started here tend to stay.

The 2-year sleep regression

Imagination arrives, and with it new fears that were not there a month ago — the dark, being alone, something under the bed. Often coincides with potty training, a new sibling or a move to a bed. Calling out, leaving the bed and early waking. Two to six weeks.

The ages people search for that are not on the list

Search for almost any month between three and twenty-four and you will find a page naming a regression at exactly that age. Most of them are one of the six above, arriving early or late in a particular baby, and it is worth saying which.

• 3 months. Usually the 4-month regression arriving early — sleep cycles mature on their own timetable, and some babies get there at eleven or twelve weeks.
• 5 months. Generally the tail of the 4-month change plus rolling. If sleep never really settled after four months, this is one long adjustment rather than two events.
• 7 months. The 8-month regression with an early crawler in it, often tangled with the three-to-two nap transition.
• 9 and 10 months. The same 8-month regression, which is why the table above gives it as a band rather than a month. Standing in the cot is the giveaway at any of them.
• 15 months. Almost always the two-to-one nap transition rather than a regression, and the fix is a nap schedule change rather than waiting it out.
• 3 years and beyond. Real, but usually driven by something nameable — a dropped nap, a move to a bed, potty training, a new sibling, or new fears at bedtime.

The practical point is that the age matters far less than what your child has just learned to do. A regression named after a month you can find on a chart still behaves like the skill underneath it.

Telling a regression from teething, illness or a leap

These overlap and frequently coincide, which is why “is it teething?” has no clean answer at 3am. What separates them is the daytime, not the night.

When two of these arrive together — and an 8-month-old crawling through their first molars is not unusual — treat the one you can act on. Teething pain and illness have their own responses; the regression underneath will still be there afterwards, and will still end on its own.

How long a sleep regression lasts

Two to six weeks is the honest answer for all of them, with the 8-month one usually at the long end and the 6- and 12-month ones at the short end. What varies is not really the length but how long it takes the underlying skill to consolidate — once a baby can crawl comfortably, the compulsion to practise at midnight goes away.

Six weeks is the number worth holding onto, because it is a decision point rather than a prediction. Past six weeks with no sign of improvement, the odds shift away from a regression and towards something else: a wake window that no longer fits, a nap that needs dropping, teething, reflux, or illness. At that point looking for a cause beats waiting it out.

Do regressions affect naps as well as nights?

Almost always, and often before the nights change. Naps are the more fragile sleep — shorter, lighter, and held together by less sleep pressure — so a baby practising a new skill loses the second half of a nap well before they start waking at 2am.

What that looks like:

• Naps collapse to one cycle. Thirty to forty-five minutes, then awake. Common in the 4-month regression, when surfacing between cycles is new.
• The first nap survives, the last one goes. Sleep pressure is highest in the morning, so the morning nap is the last to fall apart and the first to come back.
• Settling takes longer than sleeping. Twenty minutes of standing up in the cot followed by a 40-minute nap is the classic 8-month day.

The useful response is the same one that works at night, applied earlier in the day: shorten the window before the nap by about 15 minutes, and use an earlier bedtime to absorb the daytime sleep the naps are no longer producing. A day of short naps followed by a normal bedtime is what turns a regression into a genuinely bad night.

What actually helps

• Keep the routine identical. The same order of steps at bedtime is a stronger signal than anything else you can do.
• Give the new skill daylight hours. A baby who has practised rolling or standing plenty during the day is less compelled to do it at midnight.
• Shorten wake windows by 15 minutes. Developmental leaps are tiring. An overtired baby fights sleep harder.
• Pause before going in. Some night waking resolves itself in a minute or two. Some does not. Learning the difference for your baby is worth more than any rule.
• Log it. Two bad nights feel identical to two bad weeks at 3am. A record tells you which one you are actually in.

What tends to backfire

• Introducing a new sleep association you do not want permanently — a regression is the worst time to start something you will need to undo.
• Dropping a nap because it went badly for a few days. Nap transitions have their own timeline and forcing one during a regression usually makes nights worse.
• Changing approach every three nights. Nothing gets a fair test, and the inconsistency itself becomes the problem.

Should you sleep train during a regression?

Whether to sleep train at all is a decision for your family, and approaches to it differ widely. What the approaches largely agree on is the timing: a regression is a poor moment to start one.

The reasons are practical rather than ideological:

• Nothing gets a fair test. Sleep is changing on its own over the same two to six weeks, so improvement and relapse both get credited to the method rather than to the phase.
• The child is working on something else. A baby mid-way through learning to crawl or talk has less capacity for a new expectation at bedtime.
• Illness and teething often ride along. Starting something structured on top of a fever usually means abandoning it, and an abandoned attempt is harder to restart than a delayed one.

Waiting two or three weeks for a stretch of ordinary nights costs very little. What is worth doing during a regression is the part every approach shares anyway: the same bedtime routinein the same order, every night, so the signal stays constant while everything else moves.

Where tracking helps

The single most useful thing during a regression is knowing whether it is ending. Memory is unreliable when you are short on sleep — a good night after four bad ones feels like a turning point, and a bad night after four good ones feels like the whole thing restarting.

A log of night wakings and total sleep across two or three weeks shows the actual trend, and it is the only thing that reliably separates “this is ending” from “this is starting again”.

When it is not a regression

Speak to your paediatrician or health visitor if disrupted sleep comes with fever, poor feeding, weight change, breathing pauses or snoring, or if it has run well past six weeks with no sign of improving. Persistent sleep problems have causes worth ruling out, and “it is probably a regression” is not a diagnosis.