Dream Research
Why do we dream? Four serious theories and what each one gets right
There is no consensus, and anyone who tells you otherwise is selling something. But the competing explanations are more interesting than the certainty on offer, and they are not all mutually exclusive.
Every culture has an account of what dreams are for. Every generation of scientists has produced another. The honest current position is that we do not know, and that the question may be badly posed — dreaming might be several phenomena bundled under one word.
Here are the four explanations with the most substantial arguments behind them.
1. Activation-synthesis, and its descendants
Proposed by J. Allan Hobson and Robert McCarley in 1977, and initially framed as a direct challenge to psychoanalysis.
The claim was that during REM, the brainstem generates essentially random bursts of activity — PGO waves, travelling from pons to geniculate to occipital cortex. The forebrain receives this chaotic input and does what it always does: it constructs a story. Dreams are not messages. They are the cortex making sense of noise.
The theory explained a great deal about dream form: the bizarre transitions, the impossible geography, the sudden scene changes, the confabulated acceptance of absurdity.
Its weakness is dream content. If the input were random, dreams should be random. They are not. They are overwhelmingly social, they feature familiar people and places, and they return to the dreamer's actual preoccupations with a persistence that noise does not explain.
Hobson substantially revised the theory over subsequent decades, later proposing that REM constitutes a kind of protoconsciousness — a virtual reality the brain builds and rehearses, particularly important in development.
2. Memory consolidation
The best-supported function of sleep generally, and the most plausible mechanism connecting sleep to dreaming.
Sleep demonstrably improves memory. Slow-wave sleep is associated with consolidating declarative memory, apparently through hippocampal replay — sequences of neural firing recorded during learning recur during subsequent sleep, compressed in time, and are gradually transferred to cortical storage. This has been shown directly in rodents with electrode arrays.
Robert Stickgold's group added a striking human finding: people learning a task — a virtual maze, or the game Tetris — who reported dreaming about it during subsequent sleep improved substantially more than those who did not. The dream content itself predicted the learning benefit.
The complication is that dreams rarely replay episodes faithfully. Fewer than two per cent of dream reports contain a full replay of a waking event. Instead they recombine elements — the dream lag effect shows that memory fragments from about five to seven days earlier are disproportionately likely to appear.
Which suggests the process is not recording but integration: connecting new material to existing knowledge rather than filing it verbatim.
Dream content is overwhelmingly social. Across large collections of reports, the majority involve other people and interactions with them. Aggression is more common than friendliness. Misfortune outnumbers good fortune. This bias toward social difficulty is the single fact any theory of dreaming has to account for — and it is what the threat-simulation and social-simulation theories are built on.
3. Threat simulation
Antti Revonsuo's proposal is evolutionary: dreaming is a rehearsal system for dangerous situations, refined in an ancestral environment where the cost of being unprepared was death.
The supporting evidence is the content bias. Dreams contain far more threatening events than waking life does. Being chased, attacked, falling, and failing to escape are among the most commonly reported themes across cultures. Children in high-threat environments report more threatening dreams than those in safe ones.
The theory has been extended by Revonsuo and colleagues into social simulation — the idea that dreaming rehearses social bonds and conflicts, which for a hypersocial species is arguably the more relevant survival domain.
Objections: much dream threat is not adaptive rehearsal in any useful sense, since you rarely escape, and the outcomes are frequently absurd. And people with REM-suppressing medication or brainstem damage do not obviously become worse at handling danger.
4. Emotion regulation
REM sleep is associated with distinctive neurochemistry: noradrenaline, the neurotransmitter of arousal and stress, falls to its lowest levels of the entire twenty-four hours, while limbic and amygdala activity is high.
Matthew Walker's proposal is that this offers a unique opportunity — emotionally charged memories can be reprocessed and stripped of their physiological charge, without the stress response that would accompany recalling them awake. Overnight therapy, in his phrase.
Supporting evidence includes reduced amygdala reactivity to emotional images after a night with REM compared with sleep deprivation, and the observation that people with post-traumatic stress disorder show disturbed REM and fail to show this overnight reduction — the memory stays charged, and recurs.
Rosalind Cartwright's longitudinal work on people going through divorce found that those who dreamed about the ex-partner, with the dream narrative evolving over time, showed better psychological outcomes a year later than those who did not.
Why they are not necessarily rivals
REM sleep and dreaming are not the same thing — dreaming occurs in non-REM stages too, and patients with brainstem lesions abolishing REM can still report dreams. Mark Solms's work on this is among the more important challenges to the assumption that studying REM is studying dreaming.
It is entirely plausible that the brain state has several functions operating simultaneously, and that the subjective experience of dreaming is partly a by-product of processes that would happen anyway. That is a less satisfying answer than any single theory. It is also the one the evidence currently supports.
