When Do Children Start Dreaming?

Watch a newborn sleep for a few minutes and it's hard not to wonder what's going on in there. The eyelids flicker, the face twitches through expressions, the hands open and close. It looks like dreaming. And there's a fact that seems to settle it: full-term newborns spend about half their sleep in REM, the stage most associated with dreams, against roughly 20% for adults.

So babies must dream more than anyone. That's the assumption, and it's one of the more interesting places where the research disagrees with intuition.

REM is not the same thing as a dream

The cleanest way to put it: REM sleep is a brain state, and dreaming is an experience. In adults the two overlap heavily, which is why they get used interchangeably. In infants there's no reason to assume they do.

A newborn's REM looks to be doing a lot of construction work. The leading view is that it helps wire up a brain that's growing fast, giving developing circuits activity to organize around. Whether any of that comes with an experience attached is something nobody can check, because the only way to know about a dream is for someone to tell you about it. A three-month-old can't.

So the honest answer for babies is that we don't know, and the high REM number doesn't tell us. The question gets testable once children can talk, and that's where it gets surprising.

The researcher who woke up hundreds of children

The landmark work here is David Foulkes, who spent years running children through a sleep lab, starting at age three and following some of them into their teens. He'd wake them from REM sleep and ask what had been happening. His findings are collected in Children's Dreaming and the Development of Consciousness.

What he found was sparse. In G. William Domhoff's summary of the work, children aged three to five gave anything that could reasonably be called a dream on only 27% of REM awakenings. Adults woken from REM report a dream most of the time. And when the preschoolers did report something, it tended to be a single static image, often an animal, sometimes a body state like being hungry or sleeping. No plot. Very little action. Usually no self in the scene at all.

From there it builds in stages. Between five and seven, dreams start having characters and movement. Until around eight, the child mostly appears as a passive observer of their own dream rather than someone doing things in it. By roughly eleven to thirteen, Foulkes's subjects were producing dreams that looked like an adult's in frequency, length and structure.

The best predictor wasn't language

The part of Foulkes's work that has aged most interestingly is what predicted whether a child reported dreams at all.

The obvious guess is verbal ability. A child who talks better should describe dreams better. That isn't what turned up. In a cross-sectional study of five- to eight-year-olds, verbal skills didn't predict dream reporting. Visuospatial skills did, with the single best predictor being Block Design, a test where a child copies patterns using colored blocks.

That's a strange and rather wonderful result. It suggests dreaming isn't something that happens to a brain automatically during REM. It's a skill that depends on the brain's ability to build and manipulate a mental scene, and that ability has to develop before there's much of a dream to report. Foulkes's broader conclusion was that dreaming is a cognitive achievement. In his view, young children aren't dreaming poorly so much as gradually learning to dream at all.

Not everyone accepts the lab picture

This is where it would be easy to stop and declare the matter settled. It isn't.

A sleep lab is a strange place for a four-year-old. Being woken in the night by a stranger and asked to describe your inner experience is a hard task for a small child, and some researchers have argued the lab method underestimates what young children are actually dreaming.

In 2015, Piroska Sándor and colleagues in Budapest tried a different approach. Instead of a lab, they trained parents to run a structured dream interview with their four- to eight-year-olds at home, every morning, for six weeks. Their results looked quite different. In 70% of the children's dream reports, the child appeared in an active role. 82% of the reports involved movement. And 70% featured human characters, outnumbering animals.

Foulkes had already pushed back on critics of the lab method years earlier, and in fairness the two approaches have different weaknesses. Home reports are collected by parents, who might lead, embellish or simply expect more. Lab awakenings by a stranger might suppress what a shy child is willing to say. The real number for preschoolers probably sits somewhere between the two, and the field hasn't closed the gap.

What both lines of research agree on is the direction. Dreams grow more complex, more social, and more centered on the self as children get older. The disagreement is about how early it starts and how far along a four-year-old already is.

If you want to ask a child about their dreams

For a parent, the practical upshot is that young children's dream reports are fragile, and the way you ask shapes what you get.

The researchers who collected dreams at home used a structured interview for a reason. Open questions work better than leading ones. "Did you dream anything?" and "what happened next?" leave room for a child to say no or to report something small, which, if Foulkes is right, is often the true answer. "Were you scared of a monster?" hands them a story to agree with.

A short image is a real dream report. If a four-year-old says they dreamed about a dog and nothing else, that may be the whole of it, and it fits what the lab found. There's no need to push for a plot that wasn't there.

And if you keep a dream journal yourself, the research offers a small, odd comfort. Your ability to have long, populated, first-person dreams is something you built, over about a decade, alongside learning to picture and rotate things in your head. Every morning entry is a record of a skill most people never realize they had to learn.