Does sleep training harm infant attachment or cause lasting stress?
Sleep training means teaching your baby to fall asleep with less help from you — usually by putting them down awake and responding to crying in a structured way. The fear many parents voice is simple: does letting a baby cry damage the bond between you, flood them with stress hormones, or cause emotional problems later? This topic looks at what the research actually measured — not what parenting forums claim it measured.
First, the methods, because "sleep training" is not one thing:
Most of the evidence below is about graduated extinction and bedtime fading, in babies from about 6 months old. That matters — don't assume it applies to a 3-month-old. [8]
Probably not — with caveats. The evidence suggests that sleep training with graduated extinction or bedtime fading does not raise infant stress hormones, does not harm parent–child attachment, and does not cause lasting emotional or behavioral problems. In the one randomized trial that measured the stress hormone cortisol and assessed attachment directly, cortisol showed small declines rather than rises, and attachment was unaffected a year later [1]. A follow-up of a large Australian trial detected no differences between sleep-trained children and controls on any child, parent, or relationship measure at age six [2]. But the evidence isn't perfect: only one small trial measured attachment directly (attachment-specific evidence: C), unmodified "cry it out" is barely studied in trials, and nothing is known past age six.
Sleep training works, at least in the short to medium term. In the randomized trial by Gradisar and colleagues (43 infants, 6–16 months), babies assigned to graduated extinction or bedtime fading fell asleep significantly faster and woke less often than controls according to parents' daily diaries — significant effects on sleep latency, night wakings, and wake after sleep onset [1]. One caveat: the wrist-monitor (actigraphy) recordings did not show a significant group-by-time difference in wake after sleep onset or total sleep — the benefit showed up in what parents reported, not in the objective recordings [1]. A large Australian cluster trial (328 families) found the same pattern by parent report: fewer reported sleep problems at 10 and 12 months, plus lower maternal depression scores, after an intervention built on controlled comforting and camping out [3].
Infant stress did not rise — it fell. This is the finding that surprises most people. Gradisar's team measured cortisol — the body's main stress hormone — in babies' saliva in the morning and afternoon. Instead of rising, cortisol showed small-to-moderate declines in both intervention groups relative to controls. Mothers' stress also declined moderately to largely over the first month of the intervention [1]. One limitation worth knowing: saliva was sampled in the morning and afternoon, never during the crying itself — so the study can't rule out an acute cortisol rise while the baby is actually crying [1].
Attachment was unaffected. Twelve months after the intervention, Gradisar's team ran the Strange Situation procedure — the laboratory gold standard for assessing attachment, in which a trained observer classifies the parent–child bond as secure or insecure based on reunion behavior. There were no differences in secure vs. insecure attachment between the sleep-trained groups and the control group, and no differences in emotional or behavioral problems [1].
No lasting effects — positive or negative — at six years. The strongest long-term data come from Price and colleagues, who followed up 225 families (69% of the original trial) when the children were six years old. They detected no differences between the intervention and control families on any measure: children's emotional and conduct scores, sleep problems (9% vs 7%), chronic stress markers (29% vs 22%), parent–child closeness and conflict, disinhibited attachment, or parents' depression, anxiety, and stress [2]. Their conclusion was deliberately flat: "no marked long-lasting effects (positive or negative)" — which means no detected group differences, not proof that no harm is possible. That finding directly answered earlier criticism that long-term safety data were lacking [10].
No adverse effects in 52 studies. A 2006 review for the American Academy of Sleep Medicine examined 52 studies of behavioral sleep treatments in young children and concluded that behavioral interventions are effective — and that "adverse secondary effects as the result of participating in behaviorally based sleep programs were not identified in any of the studies." Some studies even found treated infants to be more secure and less irritable afterward [6]. A later meta-analysis of behavioral treatments for pediatric insomnia reached the same effectiveness conclusion [7].
What it means for the parents. This gets less attention than the baby findings, but it matters: a depleted parent is not a neutral input to a baby's life. In the large Australian trial, mothers in the sleep-intervention group scored about 1.5 points lower on a 30-point depression questionnaire at 10 and 12 months [3] — a modest shift, and the share meeting the threshold for probable depression didn't differ significantly (28 vs 35 in 100 at 10 months) [3]. But two years later, the gap had widened into something more concrete: 15 in 100 intervention mothers vs 26 in 100 control mothers scored above the community cut-point for depressive symptoms, and 4 vs 13 in 100 above the clinical cut-point [4]. Mothers also reported better sleep at 12 months — 52 in 100 vs 63 in 100 rated their sleep quality poor, and 34 vs 45 in 100 said they weren't getting enough sleep [3] — and better mental-health quality of life overall [3].
A Canadian trial that enrolled both parents (Hall et al., 2015; 235 families with infants 6–8 months) found the benefits extended beyond mothers in some respects: six weeks after a group program teaching behavioral sleep strategies, both primary and secondary caregivers reported better sleep quality and less fatigue — but depression improved only for primary caregivers, not for secondary ones [13]. And here's a curious detail: in that trial, the babies' objectively measured night waking didn't actually drop on actigraphy — what changed was parents' diaries and their perception of the problem (31% vs 60% reported two or more wakes a night; severe sleep problems 4% vs 14%) [13]. Feeling less helpless may be doing real work here, not just extra minutes of sleep.
One honest caveat: in the small Gradisar trial, mothers' stress fell but their overall mood didn't differ between groups [1] — so the mood benefits aren't universal across every study. And by age six, the earlier maternal mental-health advantage had faded to nothing, with no differences in parent depression, anxiety, or stress [2]. The parental upside looks real but time-limited: it helps while you're in the trenches, not forever. One small wrinkle in the Australian trial: intervention mothers scored slightly worse on a physical-health measure at 12 months (P=.04) — a minor finding the authors didn't make much of [3].
Prevention trials — programs given to all families, not just those with sleep problems — show much less. A 2016 meta-analysis of 9 prevention trials (1,656 families) found only a small increase in parent-reported infant night sleep, no reduction in night wakes, and a small maternal-mood benefit that may reflect publication bias [12].
What we don't know: whether sleep training changes couple conflict, relationship satisfaction, or fathers' mental health in any lasting way. Fathers are barely studied in sleep-training trials — the field knows it, and this topic won't pretend otherwise.
Overall rating: B for parental outcomes — consistent short-term benefits for maternal stress and depressive symptoms in treatment trials, but prevention trials show nothing, fathers are barely studied, and long-term couple/relationship effects are unknown.
The viral cortisol study, honestly examined. You've probably seen the claim that "studies show crying it out floods babies with cortisol." It comes from Middlemiss and colleagues (2012): 25 mother–infant pairs in a five-day residential hospital program where nurses — not the mothers — put the babies down using unmodified extinction. On the third night, the babies had stopped crying but their cortisol was still elevated after falling asleep, while the mothers' cortisol had dropped. The authors called this "asynchrony" [5].
Here's what that study can't tell you: there was no control group, so we don't know what those babies' cortisol would have done without the program. The babies were in a strange hospital ward being settled by nurses — not their mothers — which is a stressful setting in itself. And the authors never claimed long-term harm; that interpretation was added later by commentators. It's a real physiological observation in 25 infants — not evidence that sleep training damages babies.
Only one small trial measured attachment directly (attachment-specific evidence: C). Gradisar's Strange Situation finding is reassuring, but it's 43 infants from affluent, educated families. One trial, one sample, one culture. Price's five-year follow-up measured the parent–child relationship with questionnaires (closeness, conflict) rather than the Strange Situation, so the two studies complement but don't replicate each other.
Unmodified extinction is the least studied method. The "don't go back in at all" version — the one that worries parents most — barely appears in controlled trials. A 2020 systematic review noted that unmodified extinction wasn't included in any of the interventions it reviewed [9]. If this is the method you're considering, know that the evidence base is thinnest exactly where your anxiety is highest.
Nothing is known past age six. No study has followed sleep-trained children into adolescence or adulthood. The five-year data are the longest we have, and they show nothing — but "no evidence of harm at six" is not the same as "proof of no harm ever."
High-risk families are underrepresented. Trials tended to recruit stable, community samples; very preterm infants and severely depressed mothers were excluded from some studies. If your situation is complicated — a baby with medical issues, significant postpartum depression — the trial populations may not look like you, and your pediatrician's judgment matters more than these averages.
What the trials found, in concrete terms:
| Outcome | What happened | Source |
|---|---|---|
| Time to fall asleep (both methods) | Significantly faster by parent diary (P<.05); not confirmed on actigraphy | [1] |
| Night wakings (graduated extinction) | Significantly fewer by parent diary (P<.0001) | [1] |
| Wake after sleep onset (graduated extinction) | Significantly less by parent diary (P=.01); not significant on actigraphy | [1] |
| Infant cortisol | Small-to-moderate declines, not increases | [1] |
| Maternal stress | Moderate-to-large declines in the first month | [1] |
| Attachment at 12 months (Strange Situation) | No difference between groups | [1] |
| Sleep problems at age 6 | 9% (intervention) vs 7% (control) — no meaningful difference | [2] |
| Chronic stress markers at age 6 | 29% vs 22% — no significant difference | [2] |
| Any emotional, behavioral, or relationship measure at age 6 | No differences | [2] |
The honest summary: measurable short-term sleep gains for the baby, measurable stress relief for the parent, and no detectable cost to the bond or the child's development in any study that looked. "No detectable" is doing real work in that sentence — it means researchers measured these things and found nothing, not that nobody bothered to check.
This book informs your decisions; it doesn't replace professional care. When in doubt, ask a clinician who knows your baby.