In 2000, the Finnish cognitive neuroscientist Antti Revonsuo published a paper in *Behavioral and Brain Sciences* titled "The Reinterpretation of Dreams," which proposed that dreaming is not incidental to sleep but a distinct evolved function. On his account, the dreaming brain is a *threat simulator*: a low-cost, offline rehearsal system for dangerous situations, tuned by natural selection because the ancestors who mentally practiced escaping predators, evading rivals, and navigating social threat had a survival edge over those who did not.
The hypothesis is strongest as an explanation for a persistent puzzle in dream research: dream content, on average, is unpleasant. Content analyses across decades — most systematically by Calvin Hall and Robert Van de Castle, then by Domhoff and others — have found that dreams contain disproportionately more negative emotion than waking life, more misfortune than good fortune, and a striking overrepresentation of threatening events: being chased, attacked, falling, drowning, appearing unprepared, losing something important. If dreaming were random noise or wish-fulfillment, we would expect a much rosier distribution.
Revonsuo and his colleagues have accumulated several strands of supporting evidence. Cross-cultural surveys find that threat content dominates dreams in traditional and modern societies alike. Children in dangerous environments dream of threats more often than children in safer ones. Recurrent dreams — the ones that come back over and over — are disproportionately threat dreams. And the kinds of threats that show up most often (chases, animal attacks, natural disasters, hostile strangers) have an ancestral flavor rather than a modern one, which fits an evolutionary story.
The theory is influential but contested. Several alternative accounts share the basic move — dreams as offline simulation of something important — but pick a different target for the simulation. Revonsuo himself later proposed a *social simulation* extension: dreams may rehearse not just threats but the whole array of social scenarios that mattered ancestrally, including cooperation, courtship, and status conflict. Erik Hoel's more recent *overfitted brain hypothesis* argues that dreams introduce noise and novelty into the day's learning to prevent the brain from over-specializing — a kind of biological regularization.
Critics point out that many dreams are not obviously threatening, that a great deal of dream content looks more like everyday mundanity than survival rehearsal, and that even where threat content is high the simulation is often maladaptive: dreamers in threat dreams frequently freeze, fail to run, or do the wrong thing rather than practicing correct responses. Whether that counts as evidence against the theory or evidence for it (imperfect practice is still practice) is an open question.
The honest label is *live psychological theory with real empirical support* — not a proven evolutionary account. Even if the strongest version turns out to be wrong, the underlying observation that dreams tilt toward threat is a robust and interesting fact about the mind, and one worth noticing when you look back over your own recurring themes.