Why We Dream, Probably

Eugene Aserinsky was a graduate student at the University of Chicago in 1952 when he taped electrodes to his eight-year-old son Armond and watched his eye movements in sleep. The pen twitched in distinct bouts, roughly every 90 minutes. Wake him up during one of those bouts, and he reported a dream. Wake him between them, and he didn't. Aserinsky and his advisor Nathaniel Kleitman published the finding in Science in 1953. REM sleep existed, and it turned out to be a universal feature of mammalian physiology nobody had noticed.

Seven decades later, we still argue about what it's for. The dominant proposals cluster into three camps. The first is memory consolidation: during REM, the hippocampus replays the day's events and the cortex integrates them. Robert Stickgold's group at Harvard has shown that skill learning — a mirror-tracing task, a visual discrimination — improves overnight specifically in proportion to REM time. The gains survive if you keep subjects awake but sleeping; they don't survive if you suppress REM selectively with medication.

The second is synaptic downscaling, proposed by Giulio Tononi at Wisconsin. The idea is that being awake strengthens synapses indiscriminately, and the brain would saturate within days if it didn't run a net-negative maintenance cycle. Sleep, and REM in particular, prunes. Rodent cortical recordings show overall synaptic weights drop during the sleep bout. It's a plausible mechanism for why sleep deprivation makes cognition get dumber before it gets worse.

The third is emotional processing. Matthew Walker's lab at Berkeley has shown that REM sleep reduces the amygdala's response to remembered emotional stimuli the next day. Subjects who REM-sleep between a traumatic film and a recall test rate the images as less upsetting; subjects who don't, don't. The "overnight therapy" framing is controversial but keeps producing results.

None of these is mutually exclusive. A dream probably is all three at once: consolidating memory, pruning synapses, dampening affect. What muddies the story is that dreams also happen in non-REM sleep (less vivid, less narrative), and some animals — dolphins, for instance — have almost no REM and seem fine. The fur seal shows REM only when on land. The platypus REMs at rates higher than any other mammal, and we have no idea why it would need to.

A different class of evidence comes from what fails when REM goes missing. Total sleep deprivation in humans at about 64 hours produces micro-sleeps, hallucinations, and — in the extreme case of fatal familial insomnia, a prion disease — death within about a year. Rats in Allan Rechtschaffen's 1980s experiments, kept awake on a rotating disk over water, died after two to four weeks from hypermetabolism and infection. The specific lethality of REM suppression is harder to isolate in humans, but suppressing it pharmacologically with SSRIs for years seems tolerable, which argues REM isn't strictly necessary for survival, only for something subtler.

Dream content is its own puzzle. Mark Solms, a neuropsychologist who reviewed lesion studies in the 1990s, found patients with damage to the ventromedial forebrain — not the REM-generating brainstem circuits — lost dreaming while keeping REM sleep. The machinery that generates dreams is partly decoupled from the machinery that generates REM. Which means the "REM = dreams" shorthand is too tidy.

What's left is a set of working hypotheses with partial evidence and a phenomenon we all have intimate access to but cannot study from the inside. For about two hours of every night, for your entire life, your brain runs a simulation that uses almost as much energy as being awake. Evolution has kept paying the bill since mammals existed. Whatever the reason, it was worth the glucose.

The honest scientific position, at this writing, is that sleep is clearly doing something important and REM is clearly doing something specific — and nobody has a single theory that covers why the neurochemistry, the brainstem circuitry, the dream content, and the cross-species variation all line up. It's the rare well-studied biological function whose purpose remains genuinely unresolved. When someone tells you confidently what dreams are for, they are probably selling a book.