You’ve been told your baby has a “negative sleep association.” Maybe it’s feeding to sleep. Maybe it’s rocking. Maybe it’s being held, or being in a pram, or only settling with a dummy that falls out and needs replacing.

The implication is consistent: you have accidentally taught your baby the wrong way to fall asleep. The association is negative — harmful, problematic, a mistake — and now it needs to be corrected before it becomes entrenched.

But where does this concept actually come from? And does the evidence support it?

For the broader picture of what the “bad habits” narrative is and where it goes, see our complete guide. This article focuses specifically on the term “negative sleep association”: its intellectual origins, what the research actually shows, and whether the framework it belongs to is a good description of infant sleep behaviour.

What a ‘sleep association’ actually is

A sleep association is any condition present at the moment of falling asleep. Everyone has them.

Adults fall asleep with pillows, darkness, a preferred position, a particular room temperature. These conditions don’t cause sleep — but they are present at sleep onset and associated with it. On the occasional night when the pillow is missing, or the room is too bright, or you’re in an unfamiliar bed, the transition into sleep may be harder.

Babies have sleep associations too. Warmth, movement, sucking, the sound of a heartbeat, proximity to a caregiver. These are not taught. They are the conditions under which human infants have always fallen asleep, across every culture in history.

The sleep training industry divides sleep associations into two categories:

“Positive” associations are those the baby can recreate independently between sleep cycles: white noise, a sleep sack, a dark room, a comforter once old enough to have one safely. The baby wakes, finds these conditions present without parental input, and resettles.

“Negative” associations are those that require a parent to be present or to act: feeding, rocking, holding, patting. The baby wakes, these conditions are absent, and they signal for the parent to restore them.

The classification seems logical until you examine what it is actually based on. “Positive” and “negative” are not developmental categories — they are convenience categories for the parent. The classification reflects what requires adult involvement in the night, not what is developmentally appropriate for the baby.

The further irony: a baby’s most powerful biological sleep association is proximity to a caregiver. The warmth, the heartbeat, the smell, the co-regulation of physiological systems — all of these are features of the caregiving environment that human infants have relied on for sleep onset throughout our species’ existence. By the sleep association framework, the most natural sleep condition in the world is classified as negative.

The behaviourist origins

The concept of negative sleep associations has a specific intellectual genealogy.

Early twentieth-century paediatric culture — represented in Britain and America by figures like Truby King and Emmett Holt — promoted strict scheduling, limited physical contact, and early independence as the cornerstones of infant management. The values were cultural: stoicism, self-reliance, the efficient organisation of family life around the parent’s schedule rather than the baby’s needs.

In the 1920s, John B. Watson formalised the theoretical underpinning: behaviourism. In Watson’s framework, all human behaviour — including infant behaviour — is the product of conditioning through reinforcement and punishment. Nothing is instinctual. Everything is learned. The implication for infant care: if you respond to crying, you reinforce crying. If you rock a baby to sleep, you teach them that rocking is necessary for sleep. Remove the reinforcement, and the behaviour extinguishes.

Richard Ferber’s 1985 book Solve Your Child’s Sleep Problems brought the behaviourist framework into mainstream parenting culture in its modern form. The concept of the “negative sleep association” — the learned, removable, problematic condition — is the behaviourist model applied directly to bedtime. The solution Ferber offered, graduated extinction, is behaviourist intervention: remove the reinforcement on a schedule, allow the signalling behaviour to extinguish, replace the association with one the baby can reproduce alone.

The framework was genuinely useful for treating severe behavioural sleep difficulties in older children with established patterns. Its application to normal infant sleep behaviour — a three-month-old who needs feeding to sleep, a seven-month-old who wakes and calls for a parent — is considerably more contested.

What behaviourism ignores in the infant context:

  • Biology. Breast milk in the evening contains melatonin, tryptophan, and cholecystokinin — compounds that specifically induce drowsiness at the end of a feed. Feeding to sleep is not a conditioned association. It is a hormonal system doing what it was designed to do. (For more, see our guide to feeding to sleep.)
  • Attachment. Bowlby’s attachment theory — developed contemporaneously with the behaviourist sleep training tradition — established that proximity-seeking is a biological drive, not a conditioned behaviour. Babies seek caregivers because they are designed to, not because seeking has been reinforced.
  • Temperament. Infants differ constitutionally in their sensitivity, their arousal thresholds, and their capacity for self-regulation. The behaviourist model treats all babies as identical stimulus-response systems. They aren’t.
  • Developmental readiness. Self-regulation requires brain systems that are immature in infancy. The capacity for self-soothing develops as the prefrontal cortex and limbic system mature — a process that may take years [5]. The behaviourist model assumes this capacity can be installed through conditioning. The developmental neuroscience says otherwise.

What the research shows

The behaviourist framework predicts that parent-led settling creates night waking by teaching babies to signal for the conditions of their sleep onset. The research does not cleanly support this prediction.

Goodlin-Jones et al. (2001) used videosomnography — direct video observation of sleeping babies — and found that infants classified as “self-soothers” woke just as often as “signalers” at every age studied [1]. The settling method did not determine wake frequency. It determined whether the parent was notified. The association, by the behaviourist logic, should have reduced waking in self-soothers. It didn’t.

Burnham et al. (2002) found that the strongest predictor of self-soothing capacity at 12 months was high levels of quiet sleep at birth — a constitutional characteristic present before any settling technique was applied [2]. This suggests that self-soothing develops primarily through temperament and neurological maturation, not through the removal of parental involvement at sleep onset.

A 2023 review found that feeding specifically to help a baby fall asleep had no significant effect on objective sleep outcomes [3]. The strongest predictors of sleep quality were environmental: darkness and a bedtime feed for nutritional satiation. The association itself — feeding to sleep — was not what determined sleep quality.

Long-term follow-up studies of sleep training found no differences at five years between trained and untrained babies in sleep quality, emotional development, or behaviour [4]. If “negative sleep associations” created the lasting problems the framework predicts, those problems would appear at five-year follow-up. They don’t.

The nuance is worth stating clearly: there is some limited evidence that putting babies to bed awake is loosely associated with earlier independent settling. This is real evidence and shouldn’t be dismissed. But it is correlational, based on small samples, and may reflect temperament (babies who tolerate being placed awake may already be constitutionally easier to settle) rather than the causal effect of technique. It does not establish that parental involvement causes lasting sleep problems.

A better way to think about sleep associations

The “positive vs. negative” framework is not the only way to think about sleep conditions. There are more useful distinctions.

Sustainable vs. unsustainable. A sleep association is sustainable if both the baby and the parent can live with it. Feeding to sleep at four months, when everyone is happy with it, is sustainable — regardless of what the framework calls it. Feeding to sleep at fourteen months when the parent is desperate and exhausted may have become unsustainable. Sustainability is a practical, family-specific assessment. It is not determined by whether the association is “positive” or “negative.”

Dependent on one condition vs. flexible. Babies who can fall asleep with multiple cues — feeding, and rocking, and white noise, and a particular song — are less dependent on any single one. Introducing new cues alongside existing ones over time builds flexibility. This is a lower-pressure, more realistic approach than abrupt removal of all parental involvement.

What the baby needs now vs. what they’ll need later. Infant needs are not static. A four-month-old who needs feeding to sleep is not a fourteen-month-old who needs feeding to sleep. Development changes the equation continuously, without intervention. Meeting your baby’s current needs doesn’t lock in a permanent requirement.

The bottom line

The “negative sleep association” is a framework, not a fact. It was developed within a behaviourist model that has significant limitations when applied to normal infant development. It explains some things and misses others — and the things it misses are significant enough to question whether it should be the lens through which most parents understand their baby’s sleep.

Your baby needing you to fall asleep is not a negative. It is a developmental stage — one every baby passes through, on their own timeline, in their own way.

For the emotional side of this, read You’re Not Creating a Rod for Your Own Back. For the “rod for your own back” claim specifically, see Why This Advice Is Wrong.

References below.