Cry It Out: What the Evidence Does and Doesn't Say

Here is the honest summary of the cry-it-out evidence: the randomised trials we have, run on healthy babies past roughly six months, found that structured behavioural methods improve sleep and did not detect harm to stress systems, attachment, or later behaviour — including at five-year follow-up. And those trials are small, short on long-term outcomes, and mostly studied graduated extinction rather than the closed-door version people picture. So the defensible position is “no harm detected in the research we have,” not “proven safe forever,” and certainly not “proven damaging.”

I’m a working sleep consultant. I don’t have a horse in this race — I get paid either way — so this is a straight read of the literature and the arguments around it.

First, a definition problem

Half the internet arguments here are people using one phrase for four different things:

• Full extinction — down awake, no returns until morning (feeds per your pediatrician’s plan). This is what “cry it out” technically means.
• Graduated extinction — timed check-ins at increasing intervals. This is what most people actually do, and what most studies actually tested. It has its own detailed review here.
• Extinction with parental presence — you stay in the room but don’t intervene.
• “Leaving a baby to cry” — the strawman: an unfed, unwell, or newborn baby left indefinitely, which no reputable method describes and no study tested.

Almost every scary claim about cry-it-out is built by quoting research on the fourth category — chronic, unresponded-to distress, often in institutional settings — and applying it to the first three. Almost every dismissive claim is built by quoting the trials as if they’d been enormous. Both moves are sleight of hand.

What the trials actually measured

The most-cited randomised trial is the 2016 study in Pediatrics by Gradisar and colleagues, which randomised 43 infants aged six to sixteen months to graduated extinction, bedtime fading, or a sleep-education control. Both active approaches improved sleep over control, and the researchers found no adverse stress responses and no measurable effect on parent-child attachment or child emotional and behavioural outcomes at follow-up. You can read the trial itself — it’s open on the AAP’s journal site.

The longest follow-up comes from the Australian infant sleep work: 326 children from a population-based randomised trial, followed up at age six and published in Pediatrics in 2012. The conclusion was blunt — no marked lasting effects, positive or negative — with short-to-medium-term benefits for infant sleep and maternal depression symptoms.

That’s the core of it. A handful of trials, mostly modest in size, mostly on babies past six months, mostly testing graduated rather than full extinction.

What the evidence does not say

It doesn’t say full extinction is as well-evidenced as graduated. It’s less studied. Practitioners who present them as interchangeable are extrapolating.

It doesn’t say anything useful about newborns. The trials start months later. Nothing in this literature supports sleep training a young infant, and every responsible framework — mine included — puts a hard gate at roughly four months adjusted age with pediatrician sign-off. My when NOT to sleep train checklist is that gate written out.

It doesn’t rule out effects nobody measured. “No difference detected” in a trial of dozens of infants is a weaker statement than “no difference exists.” Small samples miss small effects. Anyone citing these trials as proof of safety is overreaching in the same direction as the people citing cortisol studies as proof of harm.

It doesn’t say your baby will sleep. The trials report group averages. Individual babies decline to be averages.

The cortisol argument, fairly stated

The most-circulated harm claim rests on a small 2012 study reporting that infant cortisol and maternal cortisol became desynchronised across an inpatient sleep programme — babies stopped crying while their measured stress markers stayed elevated. It’s a real study and it raises a fair question.

It also had a very small sample, no control group, and a setting that doesn’t resemble a family bedroom, which is why it can’t carry the weight either camp puts on it. What it should do is make everyone humble: measurements of infant physiology during sleep training are genuinely thin, and “the baby stopped crying” is not by itself proof of anything internal. I hold that hedge and I don’t pretend it resolves the debate.

Where I land, as a practitioner

I don’t recommend extinction to everyone, and I don’t consider it cruel. It’s the fastest method available, it’s the hardest to sit through, and it suits a specific family: one whose baby re-escalates every time a parent appears, whose adults agree, and whose pediatrician has cleared the plan. Families who need presence should be running the chair method or a fading approach instead — not a half-hearted extinction, which is the worst of both worlds.

What I will not do is tell you your choice makes you a bad parent, in either direction. Parents who sleep train are not neglectful; parents who decline are not martyrs. If a professional in this industry is scaring you toward or away from a method, note that fear is the cheapest sales tool there is.

What I’d tell a client

Read the evidence for what it is: reassuring, limited, and mostly about graduated methods on babies past six months. Clear the medical gate with your pediatrician first, then choose based on what your household can actually run for two weeks. If you choose extinction, choose it fully and briefly — days, not weeks of partial attempts. If it makes you feel sick, that’s a legitimate reason not to, and there are slower methods with honest timelines that get to the same place.

FAQ: cry it out and the evidence

Is cry it out harmful?

In the randomised research we have — on healthy babies at appropriate ages, with structured methods — no harm to attachment, stress systems, or later behaviour has been detected, including at five-year follow-up. Those studies are small and don’t cover every scenario, so the accurate phrasing is “no harm found,” not “proven harmless.”

How long is too long to let a baby cry?

There’s no evidence-based number, and any practitioner who gives you one is inventing it. What matters more: crying that escalates hour after hour, sounds different from protest, or comes with any sign of illness ends the session and starts a pediatrician conversation. Full stop.

Does cry it out work faster than gentler methods?

Generally yes — extinction is typically the quickest route, often measured in days, with graduated methods close behind. Speed is bought with intensity, and a fast method abandoned on night two is slower than a gentle method finished.

Should I use cry it out with a newborn?

No. Nothing in this literature applies to newborns, and no responsible framework sleep trains one. Newborn nights are about safe sleep, feeding, and rhythms — and any question about feeding intervals belongs with your pediatrician, not a method.