Is Sleep Training Safe? What the Research Actually Shows

"Sleep Was Going Well. What Just Happened?"

It was working. The bedtime routine, the schedule, the wake-up time. Now it's not. You're standing in the hallway at 2 a.m. wondering when your child stopped being your good sleeper.

Sleep changes constantly in childhood — every developmental leap, every growth spurt, every illness can disrupt a previously-good sleeper. The good news is that almost every sleep disruption is fixable without sleep training, in 2-6 weeks. Here is the evidence-based playbook.

What Sleep Training Actually Does

Before we discuss safety, we need to be precise about what "sleep training" means — because the term covers a wide range of approaches that are not neurologically equivalent. The methods parents typically Google about are:

Extinction (cry-it-out / CIO): Put the baby in the crib, leave the room, and do not return until morning regardless of how long or hard the baby cries. The baby eventually stops crying — typically after 45-90 minutes on the first night, decreasing over 3-7 nights until the baby no longer cries at all.

Graduated extinction (Ferber method): Put the baby in the crib, leave the room, and return at increasing intervals (3 minutes, then 5 minutes, then 10 minutes) to briefly reassure the baby without picking her up. Over several nights, the intervals increase until the baby falls asleep without the parent present.

Controlled comforting / check-and-console: Similar to Ferber but with more responsive check-ins, including brief physical comfort.

What all of these methods share is a core mechanism: the baby cries to signal distress, and the caregiver either does not respond or delays response until the baby stops signaling. The baby "learns" to fall asleep independently — but what has she actually learned? This is where the safety question becomes critical, and where Village AI's position is clear.

The Cortisol Studies Nobody Talks About

The most cited evidence in favor of sleep training safety comes from a 2012 Australian study (Hiscock et al.) and a 2016 study (Gradisar et al.), both of which concluded that sleep training produced no adverse effects on children's stress or attachment at follow-up. These studies are real, published in reputable journals, and widely cited by pro-sleep-training advocates. They are also methodologically limited in ways that the popular summaries never mention.

Here's what the favorable studies actually show — and what they don't:

Limitation 1: Cortisol was measured at the wrong time. The Gradisar study measured salivary cortisol at a single time point after sleep training was "complete" (when the baby had stopped crying). By that point, the acute cortisol spike from crying has already cleared. The study didn't measure cortisol during the extinction crying — which is when the stress response is at its peak. The absence of elevated cortisol after the baby has stopped crying doesn't mean the baby didn't experience extreme stress during the crying. It means the stress wasn't measured when it was happening.

Limitation 2: The 2012 Middlemiss study told a different story. Dr. Wendy Middlemiss at the University of North Texas measured cortisol in both mothers and babies during a sleep training program. On the first night, both mother and baby showed elevated cortisol — both were stressed. By night three, the babies had stopped crying. But their cortisol levels remained elevated — as high as the first night. The mothers' cortisol had decreased (because the baby wasn't crying, the mother perceived the situation as resolved). The baby's stress response was no longer being expressed (no crying), but it was still physiologically active. The baby had learned to stop signaling. She had not learned to stop feeling distressed.

Limitation 3: Follow-up periods are too short and measures too crude. The studies that find "no adverse effects" typically follow up at 12 months or 5 years using parent-report questionnaires and broad behavioral measures. These instruments cannot detect subtle effects on attachment quality, stress reactivity, or trust in caregiver responsiveness — the domains most likely to be affected by extinction-based sleep training. The absence of evidence (at these measurement scales, at these time points) is not evidence of absence.

Limitation 4: None of the studies have a true no-intervention control. The "control" groups in these studies typically received some form of sleep education or support — meaning the comparison is between different sleep interventions, not between sleep training and doing nothing. This makes it impossible to isolate the effect of the extinction protocol itself.

What the Attachment Research Says

The broader research on early caregiving and attachment — which spans 60 years, thousands of studies, and multiple countries — is clear on one point: caregiver responsiveness to infant distress signals is the primary builder of secure attachment. When a baby cries and the caregiver responds, the baby learns: my signals work. When I need help, help comes. The world is a place where my needs are met. This is the foundation of secure attachment — the single strongest predictor of healthy emotional development, relationship quality, and mental health across the lifespan.

Extinction-based sleep training inverts this equation. The baby cries. Nobody comes. The baby cries harder. Nobody comes. Eventually, the baby stops crying. What has she learned? Not that she can calm herself — she can't; her prefrontal cortex is barely functional. She has learned that her distress signal doesn't produce a response. The alarm system still works (cortisol is still elevated). But the behavior (crying) has been extinguished because it proved ineffective. The baby hasn't learned to self-soothe. She's learned that calling for help doesn't work.

Dr. James McKenna, a biological anthropologist at the University of Notre Dame who has spent 40 years studying infant sleep, describes extinction sleep training as "teaching the baby to give up." His research on mother-infant co-sleeping demonstrates that night waking and proximity-seeking are biologically normal, evolutionarily adaptive behaviors — not problems that need to be "trained" away. A baby who wakes at night and calls for her caregiver is a baby whose survival system is working correctly. Training her to stop calling doesn't fix a problem. It disconnects a functioning alarm.

Village AI's Position — and the Evidence Behind It

Village AI is firmly, unapologetically, evidence-based in our opposition to extinction-based sleep training methods. We do not recommend cry-it-out, Ferber, or any method that involves leaving a baby to cry without responsive caregiver presence. Our position is based on:

• The Middlemiss cortisol data showing physiological stress continues even after behavioral extinction
• The 60 years of attachment research (Ainsworth, Bowlby, Sroufe, Main) demonstrating that caregiver responsiveness builds secure attachment
• The biological anthropology research (McKenna, Ball) showing that night waking is species-normal behavior, not a disorder
• The neuroscience of infant stress (Gunnar, Lupien) showing that prolonged cortisol exposure in infancy can affect the developing stress-response system
• The absence of rigorous evidence that extinction methods are safe — the existing studies have the methodological limitations described above

This doesn't mean we dismiss the reality of parental sleep deprivation — which is genuine, measurable, and harmful. The answer to sleep deprivation is not "just endure it." It's responsive alternatives that improve sleep for both parent and baby without requiring the baby to cry without comfort. And those alternatives exist, work, and are described in the companion article to this one.

What About the Parents Who Say "It Worked"?

Many parents who used sleep training report that it "worked" — meaning the baby stopped crying and slept through the night. These reports are genuine. The baby did stop crying. The parents did sleep more. Everyone appeared happier. But "it worked" depends on how you define success. If success means "the baby no longer signals distress at night" — yes, it worked. If success means "the baby no longer experiences distress at night" — the Middlemiss cortisol data suggests otherwise. If success means "the baby learned a healthy relationship with sleep" — there's no evidence for this. If success means "the baby's trust in caregiver responsiveness was unaffected" — there's no data either way, because the studies haven't measured it with sufficient precision.

We don't judge parents who have used sleep training. Parenting is survival, and a parent who is so sleep-deprived that they're unsafe needs a solution. But the solution should be one that addresses the parent's need for sleep AND the baby's need for responsive caregiving. Those goals are not incompatible — they just require different strategies than extinction.

Tip: If you're considering sleep training because you're desperate for sleep, read the responsive alternatives guide first. Responsive settling, gradual withdrawal, co-sleeping, nursing to sleep, and sleep environment optimization can all improve sleep without requiring the baby to cry without comfort. Village AI's Mio can create a personalized sleep plan based on your baby's age, your family's sleeping arrangement, and your specific challenges — at 2am when you need it most.

When to Seek Help

If your baby's night waking is so severe that you're concerned about your own safety or functioning, talk to your pediatrician — not about sleep training, but about: whether there's a medical cause for the frequent waking (reflux, allergies, pain, ear infections, gas), whether the sleep environment can be optimized (temperature, noise, light, sleeping surface), whether co-sleeping or room-sharing might reduce waking by maintaining proximity, and whether the wake windows and bedtime are developmentally appropriate. In many cases, addressing these factors produces significant sleep improvement without any form of sleep training.

Related Village AI Guides

For deeper context on related topics, parents reading this also find these helpful: baby sleep schedule by age, how much sleep does my child need by age, why does my baby wake up at 5am and how to fix it, white noise baby sleep guide. And on the parent-side of things: bedtime routine by age newborn to school age, how to get your baby to sleep through the night without sleep training, what to do when your child wont go to sleep alone, contact naps science baby sleeps on you.