Nobody has proved why we dream. Deirdre Barrett, a dream researcher at Harvard, told Popular Science that every dream psychologist will give you a different answer. What the field does have is a handful of serious hypotheses, each with a study or two behind it.
The dreaming itself is not in doubt. BBC Bitesize puts it at around two hours a night, often about the day just gone.
more
Where dreams happen in the night
Dream science started properly in 1953. That year Eugene Aserinsky and Nathaniel Kleitman at the University of Chicago woke EEG-wired volunteers at different stages of sleep and found REM sleep. BBC Bitesize says roughly a quarter of sleep is REM, when the brain runs close to its waking level.
REM is where the strange dreams live. New Scientist reports that people dream in other stages too, but those dreams tend to be dull, flat and quickly forgotten. In REM, the hippocampus stops passing on specific memories of past events, while emotional regions become more active. The dorsolateral prefrontal cortex, which handles logic, also goes quiet, so nothing in a dream seems odd until you wake.
Non-REM dreaming turns out to matter for another reason. A 2026 PLOS Biology study woke 44 healthy adults across 196 nights of high-density EEG recording during the N2 stage of sleep. People rated their sleep as deepest when they had been either fully unconscious or deeply immersed in a dream. As the night went on and sleep pressure fell, dreams grew more immersive and sleep felt deeper, not shallower. Dreaming, on that evidence, may help you feel you slept well.
Hypothesis one: dreams rehearse danger
The Finnish neuroscientist Antti Revonsuo proposed the threat simulation theory in 2000. It treats dreaming as an old defence system that lets the brain practise spotting and escaping threats. New Scientist notes that this would explain why people so often dream of being chased or attacked.
Popular Science cites a 2005 study in support. Kurdish children exposed to war dreamed of severe threats more often than Finnish children who had not been traumatised. It also cites a 2008 study that cuts the other way. People from high-crime areas of South Africa reported fewer threatening dreams than people from low-crime parts of Wales.
Hypothesis two: dreams take the sting out of fear
A paper in Scientific Reports tested a version of Revonsuo’s idea on small-scale forager societies. The team analysed 896 dreams from 234 people, drawn from dream diaries. The dreamers were the BaYaka of the Republic of Congo, the Hadza of Tanzania, and groups in the Global North.
Both forager groups dreamed of threats often. Yet the Hadza showed few negative emotions in those dreams, and BaYaka dreams leaned towards the community. A Global North group with nightmare disorder showed the most negative emotion, and Canadian students dreaming during the COVID-19 pandemic showed the most anxiety. The authors read this as support for dreams regulating emotion, by pairing a threat with a setting that is not frightening.
Other work points the same way. Popular Science describes a study of recently divorced people with depression. Those who dreamed of their former spouse were more likely to show better mood a year later. BBC Bitesize says a group of American scientists described dreaming in 2011 as a kind of overnight therapy.
Hypothesis three: dreams file memories
BBC Bitesize describes a rival view in which dreaming is when the brain moves memories into long-term storage and prunes connections it does not need. Popular Science frames the dream as a by-product of that work. The hippocampus and neocortex blend new memories with old ones, and the mash-up is what you experience.
Hypothesis four: dreams search for new connections
Two sleep researchers set out a more specific version in Nautilus. Their model, NEXTUP, says a dream takes a new memory from the day and looks for weakly linked older ones. The result is a narrative that tests associations the waking brain would never try. In a 1999 study, one of the pair found that people woken from REM sleep favoured weak word associations more than they did by day.
A 2026 experiment gives this idea some support. Karen Konkoly and Ken Paller at Northwestern University played puzzle-linked soundtracks to 20 sleepers during REM sleep, Live Science reports. Three quarters of them dreamed about the unsolved puzzles. The next day, they solved 42% of the puzzles that had appeared in their dreams and 17% of those that had not. The volunteers may have dreamed about the puzzles they found most interesting, so the result is not proof.
Why the question is so hard to settle
Antonio Zadra of the University of Montreal told Quanta Magazine that researchers never study dreams directly. They study what people report once they wake. He added that there is no agreed definition of a dream at all. Some researchers mean a vivid story out of REM sleep, and others mean any thought or image during sleep.
Barrett suggests the question itself may be wrong. Nobody asks what thinking is for, she told Popular Science, and dreaming may serve as many purposes.
