The average person dreams for roughly two hours per night, mostly during REM sleep, the stage characterized by rapid eye movement, elevated brain activity, and near-total muscle paralysis. Over a 70-year lifespan, that adds up to approximately 50,000 hours, or around six years, spent in states of vivid internal experience that disappear almost entirely on waking. Something this biologically expensive tends to have a function. For dreaming, the function remains genuinely unclear.
This is not a soft uncertainty of the “we haven’t quite worked it out” variety. It is a fundamental disagreement among researchers about what question to even be asking.
The most influential theory from the 20th century, Sigmund Freud’s view that dreams are disguised wish-fulfillment and a route to repressed unconscious content, is not seriously defended in contemporary neuroscience. The mechanisms Freud proposed have no anatomical basis, and attempts to test his predictions experimentally have not supported them.
What current research has established is that something important happens to memory and emotion during REM sleep. Studies consistently show that memories consolidate during sleep: information learned before sleep is better retained afterward, and sleep deprivation impairs this consolidation. REM sleep specifically appears important for procedural skills and for processing emotionally significant experiences. A 2002 paper by Matthew Walker and colleagues at Harvard showed that subjects who were allowed to sleep but had their REM sleep disrupted showed worse emotional memory processing than those who slept normally.
Matthew Walker’s more recent work (his 2017 book “Why We Sleep” popularized many of these ideas, though some specific claims in it have been disputed by other researchers) argues that REM sleep essentially divorces emotional memories from their raw affective charge, allowing recall of events without re-experiencing their full emotional intensity. On this view, dreaming is a byproduct of a system that processes and contextualizes emotional experience.
Other researchers argue that dreaming is fundamentally about simulation: the sleeping brain runs predictive models of social and physical scenarios, rehearsing responses to potential threats. People who are anxious or facing life challenges dream more vividly and more often about threatening scenarios.
There is also the activation-synthesis hypothesis, proposed by Hobson and McCarley in 1977, which holds that the content of dreams is essentially meaningless: the brainstem generates random activation signals during REM, and the cortex tries to construct a narrative from them. On this view, dreams are the result of pattern-seeking machinery trying to make sense of noise.
None of these frameworks has produced predictions that other researchers have consistently confirmed. The challenge is methodological: studying the subjective content of dreams requires subjects to report them, and dream reports are unreliable, fragmentary, and vary enormously with how and when subjects are woken. fMRI can show which brain regions are active during REM sleep, but what those activations mean for conscious experience is unclear.
The honest position is that we know dreaming is a real phenomenon, that REM sleep matters for cognitive and emotional functioning, and that disrupting it produces measurable harms. We do not know whether the experience of dreaming is what does the work, or whether dreaming is simply a byproduct of whatever the sleeping brain is actually doing.
Six years of your life happen in that gap.
