Why do we dream?

In 1953 Aserinsky and Kleitman reported regularly recurring periods of rapid eye movement during sleep, and early studies found that most awakenings from those periods produced vivid dream recall. That observation gave dreaming, known until then only from what sleepers said afterwards, an electrical signature to set beside the report.

REM sleep is characterised by cortical activity resembling waking, near-total skeletal muscle paralysis, and distinctive neurochemistry: noradrenaline and serotonin drop away while acetylcholine remains high.

Sleep deprivation degrades cognition, mood and immune function, and sustained deprivation is fatal in animal models. Sleep plays a well-documented role in consolidating memory, with slow-wave and REM sleep both thought to contribute, though recent replication attempts suggest the benefit is smaller and less robust than once assumed.

Dream reports can be obtained from non-REM sleep as well, which undercuts any theory that identifies dreaming with REM physiology alone. The two are strongly associated, not identical.

The activation-synthesis model proposed that the forebrain, switched on automatically during REM sleep, builds the dream by matching signals generated in brainstem circuits against what is stored in memory. On that account the content is a by-product of the state rather than a message. The model has been revised substantially since 1977, and the argument over it is unresolved.

Functional accounts point to emotional regulation, memory consolidation and rehearsal of threatening situations. Each is supported by some evidence and contradicted by other evidence.

Dream reports share many characteristics across cultures, and some themes, such as flying or teeth falling out, are widely shared. Whether those patterns reflect a functional process or shared features of how brains generate narrative under these conditions is not settled.

While dreaming, critical judgement is largely absent: impossible events are accepted without surprise, and holding contradictory beliefs at once is common. Reduced activity in prefrontal and parietal regions during REM sleep may be part of the reason, but no account explains why this specific failure of evaluation accompanies this specific state, or what restores awareness in lucid dreaming, when the dreamer becomes aware of dreaming.

Monoamine oxidase inhibitors, taken at ordinary doses for depression, can suppress REM sleep completely for as long as the treatment lasts, which may be months or years, and Siegel’s review reports that dreams go with it. No major memory deficit has been reported in the people taking them. This sits awkwardly with strong functional accounts. One reply is that the drugs raise serotonin and catecholamine levels, which could make up for the lost REM sleep, and the two have never been fully reconciled.

Dreams may be a simulator: a safe environment in which to rehearse threatening situations. Revonsuo, who proposed this, reads normative dream content, nightmares and the dreams of traumatised people as evidence that dreaming is specialised for simulating threat, and his group reports strong support from later tests. A test by another group found the reverse of what it predicts: people in a high-crime area of South Africa, with more recent exposure to life-threatening events, reported fewer threat dreams than people in Wales. Where the content does fit rehearsal, it fits a great deal else as well.

Dreaming may be what the reorganisation of memory feels like from the inside, fragments being replayed, recombined and integrated, experienced as narrative because narrative is what this brain produces when left running.

The measurements that connect the two are thin. Replay is recorded in bursts of a few hundred milliseconds, and only one or two per cent of dream reports contain a genuinely episodic memory from the preceding day. People who dreamed about a maze task improved on it more overnight, in a replication comparing eight such dreamers with nine others, while other attempts found no link at all.

Dreaming may have no function of its own, and happen whenever a sleeping brain is active enough and cut off from the world. Domhoff, in the neurocognitive tradition, calls dreaming a spandrel, a by-product of the evolution of sleep and consciousness. This is a legitimate hypothesis rather than a failure of imagination, and it is difficult to test.