Research

Sleep training and stress hormones: reading the studies fairly

What the two most cited cortisol studies of sleep training actually measured, what each found, and the limits of both, explained plainly for parents.

The Ferber team September 29, 2026 6 min read

Here is what usually happens, with a research note citing Price et al., Pediatrics, 2012.

Somewhere in your reading, you have probably met a version of this claim: a baby left to cry stops crying, but their stress hormones stay high. It is an unsettling idea, and it can stop a tired parent in their tracks.

Here is a fair summary. That claim comes mainly from one small 2012 study of 25 babies in an inpatient sleep program using full cry it out, with no comparison group. A later randomized trial of graduated extinction, the approach the Ferber method uses, found babies’ cortisol went slightly down compared with a control group. Both studies are small and measured different things, so neither settles the question alone. Read together, they do not show that the Ferber method raises babies’ stress hormones.

What cortisol is, and why researchers measure it

Cortisol is a hormone the body releases in response to stress, and it also rises and falls naturally across the day. Researchers can measure it in saliva, which makes it practical to collect from babies.

It is a useful signal, but a tricky one. A single reading depends on the time of day, the baby’s routine, and what happened just before. That is why when and how a study took its samples matters as much as the headline result.

The Middlemiss 2012 study: what it did and found

This is the study behind the “quiet but still stressed” claim (Middlemiss et al., 2012).

Who and where. 25 infants aged 4 to 10 months (11 boys) took part in a 5-day inpatient sleep training program, run in a clinical unit instead of at home. Babies learned to self-settle through extinction of crying, the research term for full cry it out, during the transition to sleep.

What was measured. Saliva was sampled from mothers and babies at the start of the babies’ night-time sleep and again after the babies fell asleep, on two program days. It was later tested for cortisol.

What it found. On the first day, mothers’ and babies’ cortisol levels moved together. When babies showed distress at the sleep transition, mothers’ and babies’ cortisol was again positively linked. On the third day, babies no longer showed behavioral distress at the sleep transition, but their cortisol levels were raised. Without their baby’s distress cue, mothers’ cortisol decreased. The authors described this as asynchrony: mother and baby were no longer in step.

That day-3 result is where the idea comes from.

The limits of the Middlemiss study

Presenting this study fairly means giving its finding and its limits together.

It was small: 25 babies.
It had no comparison group. There was no group of babies who were not sleep trained, so it is hard to know what their cortisol would have done in the same unfamiliar setting over the same days.
It studied full extinction in an inpatient setting. That is a different method, in a very different environment, from graduated extinction at home.
It measured two days, around bedtime. The published abstract describes no follow-up after the program.
The question it asked was about synchrony between mother and baby, which is a narrower question than whether sleep training is stressful or harmful.

The study prompted a published letter from three researchers behind the infant sleep trials discussed above, titled “Let’s help parents help themselves: a letter to the editor supporting the safety of behavioural sleep techniques” (Price, Hiscock and Gradisar, 2013). The original authors published a response (Middlemiss et al., 2013). The exchange is a reminder that researchers themselves read this study differently.

The Gradisar 2016 trial: what it did and found

The other key study is a randomized controlled trial published in Pediatrics (Gradisar et al., 2016).

Who and where. 43 infants aged 6 to 16 months were randomly assigned to graduated extinction (14 babies), bedtime fading (15) or a sleep-education control group (14). Families did the programs at home.

What was measured. Sleep was tracked with parent diaries and actigraphy, a movement-tracking wristband. Infant stress was measured with morning and afternoon salivary cortisol samples. Mothers reported their own mood and stress. Twelve months later, mothers reported on their children’s emotional and behavioral problems, and mother-child pairs took part in the strange situation procedure, a standard test of attachment.

What it found.

Graduated extinction led to large decreases in the time babies took to fall asleep, the number of night wakings and time awake during the night. Bedtime fading also led to a large decrease in the time babies took to fall asleep.
Salivary cortisol showed small-to-moderate declines in the graduated extinction and bedtime fading groups compared with controls.
Mothers’ stress showed small-to-moderate decreases in both the graduated extinction and bedtime fading groups over the first month, with no differences in mood.
At 12 months, there were no significant differences in emotional and behavioral problems or in secure versus insecure attachment.

The limits of the Gradisar trial

The same fairness applies here.

It was also small: about 14 or 15 babies per group.
Cortisol was sampled in the morning and afternoon, not right around bedtime, when the crying happens. It tells you about babies’ daytime stress levels across the program, which is a different window from the one Middlemiss measured.
The follow-up was one year. Longer-term questions need other studies.

Reading the two side by side

Middlemiss looked closely at the bedtime moment, with full extinction, in an inpatient program, without a control group. Gradisar looked at daytime cortisol across a home program of graduated extinction, against a control group. They differ in method, setting, timing and design, so they can both be accurate without contradicting each other.

What can be said with reasonable confidence: the one controlled trial of graduated extinction that measured cortisol found small-to-moderate declines. And the longest follow-up, which tracked children to age 6, found children’s outcomes were no different whether or not families had sleep trained, including a measure of chronic stress (29% vs 22%, not a significant difference) (Price et al., 2012). Our research page and our article on long-term effects go into that follow-up.

What cannot be said: that any method is stress-free, or that the question is closed. Crying is stressful for babies and for parents in the moment, and it is fine to find that hard. If it feels unbearable, our page on parent anxiety and the method comparison may help you find an approach you can live with.

A next step for tonight

If cortisol is the worry holding you back, write it down in one sentence and bring it to your pediatrician at the next visit, along with the two study names above. A worry that has been talked through is easier to weigh than one that circles at 2 a.m.

If you go ahead with the Ferber method, the Ferber app runs each wait, alerts you at check-in time and keeps a nightly history, so you can see how your baby’s nights are changing.

Common questions

Does the Ferber method raise a baby's cortisol?

The best direct evidence suggests not. In a 2016 randomized trial of 43 babies aged 6 to 16 months, salivary cortisol showed small-to-moderate declines with graduated extinction, the approach the Ferber method uses, compared with a control group. The study was small, and cortisol was sampled in the morning and afternoon.

What did the Middlemiss 2012 cortisol study find?

It followed 25 babies aged 4 to 10 months through a 5-day inpatient sleep program using full extinction. On day 3, babies no longer showed distress at bedtime, but their cortisol was high, while their mothers' cortisol had dropped. It had no comparison group and studied full extinction in an inpatient setting.

Do these cortisol studies contradict each other?

Not directly. They studied different methods, in different settings, with different designs and sampling times. Middlemiss measured full extinction in an inpatient program around bedtime without a control group; Gradisar measured graduated extinction at home, with morning and afternoon samples, against a control group.

Is there evidence of long-term effects from sleep training stress?

The longest randomized follow-up, published in Pediatrics in 2012, found children's outcomes at age 6 were no different whether or not families had sleep trained, including a measure of chronic stress. The authors concluded the techniques have no marked long-lasting effects, positive or negative.

Sources

Written by The Ferber team. Educational information, not medical advice. If you are worried about your child’s health, feeding or growth, talk to your pediatrician.

All articles