What Do Babies Dream About? The Science of Infant Dreaming

By Ron van Cann · May 2026 · 6 min read

A sleeping newborn smiles — then grimaces — then makes small sucking movements with their mouth, their limbs twitching, their eyes moving beneath still-closed lids. Something is clearly happening in there. The question of what has fascinated parents and scientists in equal measure.

The science of infant dreaming is both simpler and more surprising than most people expect.

---

The REM Paradox

If you assume that dreaming happens during REM sleep — a reasonable starting assumption — then newborns should dream more than anyone.

Newborns spend approximately 50% of their sleep time in REM, or what sleep researchers call "active sleep" in infants (the label changes in recognition that infant REM is not quite the same as adult REM). For adults, the proportion is around 20–25%.

Premature babies sleep even more actively. Infants born at 28–30 weeks of gestation may spend 80% or more of their sleep time in active sleep. The proportion decreases across the first year of life, then continues decreasing through childhood until it stabilises near adult levels.

If dreaming were simply proportional to REM, infants would be the most prolific dreamers in the species. The reality is more complicated.

---

What Infant Active Sleep Looks Like

The active sleep of infants is visually distinctive. During these periods, you may observe:

Rapid eye movements — the defining feature that gives REM its name, present in infant active sleep from birth

Facial expressions — smiles, grimaces, frowns, sucking movements; the expressions cycle through in apparent response to whatever internal activity is occurring

Limb twitching — small jerks and movements of the arms, hands, legs, and feet; brief and repetitive

Irregular breathing — compared to the slow, regular respiration of quiet (NREM) sleep

Vocalisations — small sounds, whimpers, brief cries, noises

These behaviours made early researchers confident that babies were experiencing rich dream lives. The movements, expressions, and apparent responsiveness to something suggested active inner experience.

---

The Developmental Complication

The assumption that infant REM means infant dreaming — in the adult narrative sense — runs into a significant problem from developmental psychology.

David Foulkes conducted the definitive laboratory studies of children's dreaming, waking children during confirmed REM sleep and asking them to report their dreams. His findings, published in Children's Dreaming and the Development of Consciousness (1999), established a developmental timeline that challenges the infant-dreaming assumption:

• Under 5 years old: children woken from REM sleep in the laboratory report almost nothing. Occasionally a simple static image (a balloon, an animal). No narrative, no movement, no social scenarios. The absence is striking.
• Ages 5–7: simple, static imagery begins to appear. Dreams more like drawings than movies.
• Ages 7–8: self as actor begins to appear in dreams; some narrative emerges.
• Ages 9–11: dreams begin to resemble adult dreams, with narrative complexity, social interaction, and emotional content.

The implication is important: the capacity for narrative dreaming develops alongside the cognitive capacities that support it — self-representation, narrative construction, autobiographical memory. These capacities are absent in infants and only developing slowly in toddlers and young children.

If 3-year-olds, with all their linguistic and cognitive development relative to infants, report almost nothing from REM sleep, what is happening during the vast stretches of infant REM? Not narrative dreaming, almost certainly. Something else.

---

The Sensorimotor Calibration Hypothesis

The leading current explanation for infant REM sleep's function is not dreaming but sensorimotor calibration: the use of active sleep to build, test, and integrate the developing sensorimotor maps of the infant brain.

Research by Mark Blumberg and colleagues at the University of Iowa has investigated the twitching movements of sleeping infants in detail. The findings are striking:

Infant twitches are not random. The limb movements during active sleep are organized and systematic — they activate motor cortex circuits in patterns that are consistent with sensorimotor map development. Specific movements correspond to specific activations in specific cortical regions.

Twitches precede the corresponding movements in waking. Some movements that appear first in active sleep subsequently appear as waking voluntary movements — suggesting that active sleep is "practicing" movements before they are available for waking use.

The function parallels known functions of adult REM. Just as adult REM processes and consolidates daytime learning, infant active sleep appears to process and consolidate the developmental "learning" occurring in the rapidly forming infant nervous system — in this case, not episodic memories but sensorimotor maps.

The extraordinary proportion of active sleep in premature infants (80%+) supports this hypothesis: the more developing there is to do, the more calibration is needed. As neural hardware is established and sensorimotor maps are built, the proportion of active sleep needed decreases — hence the steady decline from infancy through childhood.

---

Fetal Active Sleep

The REM-like state appears even before birth.

Fetal active sleep — identified by rapid eye movements and organised motor activity — is present from approximately 23–28 weeks of gestation. The fetal brain is generating this state well before it has the sensory experience, cognitive capacity, or nervous system maturity for anything resembling narrative dreaming.

The fetal experience that could be dreamed is limited: muffled sounds from outside the womb, proprioceptive sensations from movement within the amniotic environment, the rhythmic signals of the maternal cardiovascular and respiratory systems. The fetus has essentially no visual experience (the womb is dark) and no independent motor experience against gravity.

Whatever fetal active sleep represents, it is consistent with the sensorimotor calibration hypothesis: the developing nervous system generating and testing its own activations offline, building the foundational neural architecture that will be needed for life outside the womb.

The continuity from fetal active sleep → newborn active sleep → infant REM → adult REM is one of the developmental threads that connects the prenatal nervous system to the fully formed dreaming brain.

---

If Infants Do Experience Something

The evidence strongly suggests that infants are not having narrative dreams. But it does not conclusively establish that infant active sleep is experienced as nothing at all.

If infant active sleep is accompanied by any subjective experience, what would it likely be?

Sensory fragments, not narrative. An infant's experiential history consists almost entirely of sensory experience — the warmth and smell of caregivers, sucking and satiation, the sounds of familiar voices, tactile sensations. If these are represented in active sleep, the experience would be sensory and relational, not narrative.

Emotional states. The emotional circuitry (amygdala, limbic system) is among the earlier-developing neural systems, and infants clearly have emotional states. Active sleep may involve the cycling of emotional states — comfort, distress, hunger-satiation — without the cognitive structure to organise them into a narrative.

Motor sensations. Given the sensorimotor calibration function, the experience (if any) may centre on motor activation — the internal "felt" sense of movement that the calibration process generates.

None of this is "dreaming" in the way adults experience it. It may be more like the feeling-states and sensory textures that underlie experience before the cognitive architecture for narrative arrives.

---

When Does Dreaming Begin?

Given Foulkes's developmental evidence, the emergence of genuine narrative dreaming — dreaming that the dreamer can report and that contains the self as an actor in a developing scenario — appears to begin in earnest around ages 7–8, with precursors appearing from around 5 years.

This timeline coincides with the development of the cognitive capacities that dreaming requires: autobiographical self-representation, narrative construction, the kind of episodic memory that can be drawn on to build a dream scenario.

The infant's remarkable proportion of active sleep is not wasted — it serves the foundational developmental work that makes later narrative dreaming possible. But the dreaming itself is a later arrival.

---

The Hypnos app supports tracking dreams across the adult lifespan — from the moment narrative dreaming is established through all the stages of adult life in which it evolves.