In Behavioral and Brain Sciences, 2000, Revonsuo proposed that dreaming is not noise but rehearsal: the brain selects threatening waking material and simulates it repeatedly in the safe hallucination of sleep, practicing perception and avoidance. The amygdala, hyperactive in REM, supplies the engine. Tests: 66 percent of recurrent dreams contain threats aimed at the dreamer; traumatized Kurdish children dreamt more threats than Finnish children; but South Africans in high-crime areas dreamt fewer than the Welsh, and under 15 percent of threats are realistic survival scenarios, and the dreamer rarely escapes. Reviews call the support strong and the theory live, not settled.
The theory's dark edge: if rehearsal works, nightmares are functional, and recurring nightmares are drills that never pass inspection. Trauma research finds the loop, the threat simulation system stuck replaying the unescapable, which looks less like training and more like a scar that trains itself. The competing models, memory consolidation, emotional regulation, Solms's seeking system, each explain the slice TST cannot: the flying dreams, the absurd ones, the reunion with the dead.
The claim: TST inverts the dream's ownership, you do not have the dream, the simulation has you, conscripting consciousness nightly into a prehistoric training program written by the Pleistocene, and the 3 a.m. anxiety loop is a saber-tooth drill running on hardware that now faces emails. Verdict: the evolutionary mismatch reading is mainstream. The archive notes the echo of its simulation file: whether the simulator is evolution or a server farm, the experience from inside is identical, being run. Zhuangzi's butterfly would ask which side of the drill you are on. The amygdala does not answer.
If dreams are drills, who schedules the fire alarm, and why is it always at 3 a.m.?