Sleep Training: What the Evidence Actually Says

Sleep training is one of those topics where everyone you talk to has a strong opinion and most of those opinions are based on something other than the research. What follows is the research.

This is not an argument for or against sleep training. It is a summary of what the available evidence shows about the methods, the outcomes, and the limitations of the science. What you do with it is a family decision, and no study can make that decision for you.

The methods

There are four broad categories. Each one works differently, takes a different amount of time, and feels different to the parent.

Graduated extinction (Ferber)

You put the baby down awake, leave, and return at increasing intervals (3 minutes, then 5, then 10) to offer brief reassurance without picking them up. The intervals lengthen nightly. The check-ins serve two purposes: they reassure the baby that you have not disappeared, and they reassure you that the baby is safe. The check-in is brief, about 1-2 minutes. You speak to them, you may pat briefly, but you do not pick up and you do not feed to sleep.

This typically produces results within 3-7 nights [3]. The second or third night is often the hardest, due to a well-documented phenomenon called the extinction burst: crying increases briefly before it decreases, because the baby is testing whether the old rules still apply.

Full extinction (cry it out)

You put the baby down awake, leave, and do not return until morning (or until a pre-set feeding time). This is the fastest method in terms of results, often showing improvement within 2-3 nights, but it is also the hardest for most parents to sustain [1].

Despite the name, this method does not mean ignoring a baby who is in distress from illness, hunger, or a safety issue. It means not intervening in the process of falling asleep once you have confirmed the baby is safe, fed, and comfortable.

Bedtime fading

You temporarily shift bedtime later, to a time when the baby falls asleep quickly, then gradually move it earlier. This avoids extended crying because the baby is genuinely tired at the later bedtime. It takes longer, typically 2-3 weeks, but involves less protest [3].

The logic: if the baby currently takes 45 minutes to fall asleep at a 7 pm bedtime, start at 8 pm instead. The baby falls asleep quickly because sleep pressure is high. Move bedtime earlier by 15 minutes every 2-3 nights until you reach the target.

Gentle and responsive methods

A range of approaches that involve gradual parental withdrawal: sitting beside the crib and slowly moving your chair toward the door over days, picking up to soothe and putting back down, or a slow transition from rocking to patting to presence only. These take the longest (often 2-4 weeks) and have less clinical trial data than extinction-based methods, though observational evidence supports their effectiveness [6].

The Hall et al. trial of 235 families showed that a non-extinction-based cognitive-behavioral intervention reduced night wakings, with effects maintained at 6-week follow-up [6]. This is evidence that formal cry-based methods are not the only option.

What the evidence shows

The Mindell et al. 2006 meta-analysis examined 52 studies and found strong evidence that extinction and graduated extinction effectively reduce night wakings and improve bedtime behavior [1]. The effect sizes are meaningful: most families see a reduction from multiple night wakings to one or zero.

Hiscock et al. conducted a cluster RCT of 328 infants and found that behavioral sleep intervention reduced both infant sleep problems and maternal depression [4]. This is an important finding because parental mental health is often the missing variable in sleep training discussions. A parent who is not sleeping is a parent whose capacity for responsive caregiving during the day is diminished.

The question most parents actually care about is harm. The two most important studies here are:

Price et al. (2012): followed 225 families for five years after a sleep training intervention. At age 6, there were no measurable differences between trained and untrained children on emotional health, behavior, sleep quality, cortisol, or parent-child attachment [2]. This is the longest follow-up study available, and its findings are straightforward: the trained and untrained groups were indistinguishable.

Gradisar et al. (2016): a randomized controlled trial of 43 infants. Salivary cortisol was measured during and after sleep training. There were no significant differences in cortisol between graduated extinction, bedtime fading, and control groups at any point, including 12-month follow-up [3]. The cortisol data is relevant because a common concern is that sleep training elevates stress hormones in babies. This study measured them directly and found no elevation.

These are reassuring findings, with a caveat: the sample sizes are modest. Sleep training research is genuinely difficult to conduct at scale because you cannot blind parents to whether they are using a method, and attrition is high because sleep-deprived parents drop out of studies. The evidence is the best we have, not the last word.

The critiques

Blunden and Baills (2011) raised legitimate methodological concerns about the existing evidence: reliance on parent-reported outcomes, variable definitions of “sleep through the night,” and limited diversity in study populations [5]. These are fair points. They do not mean the evidence is wrong; they mean it is imperfect, which is different.

The argument that sleep training teaches babies their cries will not be answered is a theoretical concern, not an empirical finding. The follow-up data through age 5-6 does not support it. That said, the follow-up data comes from families who chose to participate in research, and the absence of detected harm is not proof that no child was ever harmed. The evidence is reassuring, not absolute. Acknowledging uncertainty is not the same as recommending against something; it is being honest about what we know and what we do not.

There is also a selection bias in who participates in sleep training trials. Families who are willing to randomize into a sleep training condition may differ from families who would never consider it. The research tells us about families who were open to the method; it cannot tell us about families who were not.

Choosing a method

The research does not identify a single best method. What it does suggest:

• Extinction-based methods work fastest but involve more crying in the short term.
• Gradual methods take longer but involve less parental distress.
• Consistency matters more than which method you choose. Switching methods mid-course resets the learning process and extends the period of disruption for everyone.
• Parental mental health matters. A method that leaves a parent in distress every night may be worse for the family than one that takes longer but is sustainable [4].

The best approach is the one your family can do consistently for at least a week. If you cannot sustain it, it is the wrong approach for you, regardless of what the evidence says about its effectiveness in aggregate.

It is also worth noting that sleep training does not have to be all or nothing. Some families use a gradual method for bedtime but allow comfort at night wakings. Others sleep-train for falling asleep at bedtime but keep one night feed. The methods described in clinical trials are presented as distinct protocols, but real families adapt them. The core principle across all of them is the same: give the baby the opportunity to practice falling asleep without the conditions they have relied on, and do it consistently enough that the new pattern can take hold.

What to expect

Night 1-3 (extinction-based): The hardest stretch. Crying may last 20-45 minutes. The extinction burst, if it happens, peaks around night 2 or 3.

Night 4-7 (extinction-based): Crying at bedtime typically drops to under 10 minutes. Night wakings begin to resolve.

Week 1-2 (gentle methods): Some improvement, often modest. The baby is learning the new pattern. Less crying per night, but progress is slower.

Week 2-4 (gentle methods): More visible progress. Bedtime protests decrease; self-settling improves.

If you see no improvement after a week of consistent application (for extinction methods) or three weeks (for gentle methods), consider whether something else is going on: illness, teething, hunger, or a method that is not the right fit for your baby’s temperament.

A note for tonight

Whatever you decide, you are not failing your child by considering this. You are also not failing them by deciding against it. The research gives you options; your family decides which one fits. And if the answer is “not tonight,” that is a perfectly reasonable answer.

The only choice that the evidence does not support is guilt. The guilt that comes from sleep training, and the guilt that comes from not sleep training, are both manufactured by a culture that has turned an evidence-based parenting decision into a moral test. It is not one. It is a decision about how your family sleeps, made by the people who live in your house.

If you’re concerned about your baby’s sleep or health, speak with your pediatrician or family doctor. The American Academy of Pediatrics has sleep safety guidance at healthychildren.org. If your own sleep deprivation is affecting your daily functioning or mental health, that is also worth raising with your doctor.