How dreams are studied in the laboratory

Dreaming is a private experience with no external trace, which makes it an awkward object of study. Everything known about it comes from a small set of methods, each with a specific reach and a specific blind spot. Knowing what they are makes it much easier to judge any claim about dreams — including the ones in this section.

Polysomnography is the foundation. An overnight recording combines EEG for brain activity, electro-oculography for eye movements, and electromyography for muscle tone, usually with breathing, heart rate and oxygen measures alongside. Those three core channels are what allow sleep to be scored into stages: the slow waves of N3, the near-waking EEG of REM, the eye movements that give REM its name, and the muscle atonia that distinguishes it. Every claim in this section about which stage something happens in ultimately rests on this.

But polysomnography cannot see a dream. It shows the state, not the experience. To get the experience, researchers have to wake people up and ask — which is the second core method and the one everything else depends on. Aserinsky and Kleitman's original insight was exactly this: waking sleepers during eye-movement periods produced vivid dream reports far more often than waking them at other times. Serial awakenings extend the technique, waking the same person repeatedly across a night to sample what is going on at different points, and it was this approach that established that dreaming also occurs outside REM, in the work described in this section's article on REM sleep.

Outside the laboratory, dream diaries collect far more material, from ordinary sleep in ordinary beds, over long periods. Their weakness is obvious — they record only what people recall and choose to write — but their volume and ecological validity are unmatched, and long individual series can span years. To turn that material into data rather than anecdote, researchers use content analysis, most influentially the Hall and Van de Castle coding system: explicit rules for counting characters, settings, social interactions, emotions, aggression and misfortune, so that different coders score the same report the same way. That is what makes claims like "most dream content is mundane" quantitative rather than impressionistic.

Two modern landmarks are worth naming. Francesca Siclari and colleagues used high-density EEG with serial awakenings and identified a posterior cortical region — a "hot zone" — whose activity tracked whether a sleeper reported dreaming, in both REM and non-REM sleep. That is a significant result because it decouples dreaming from sleep stage: it points to a neural correlate of dreaming itself rather than of the state it usually occurs in, and the researchers reported being able to predict whether a sleeper would report a dream before waking them.

The second is decoding. Yukiyasu Kamitani's group combined fMRI with machine learning to identify broad categories of imagery in sleep-onset dreams — training a classifier on brain activity while participants viewed images awake, then applying it to activity during sleep and comparing its output with the participants' own reports. It works at the level of general categories, not scenes or narratives, and it is a long way from anything resembling recording a dream. But it is a genuine demonstration that dream content leaves a readable signature.

The limitations are the honest close, and they apply to everything above. Every dream report is a memory of an experience, given by someone who has just been woken, filtered through the language available to describe something that resists description — and, as the article on forgetting dreams explains, dream memory is fragile and reconstructive. The laboratory itself changes what it measures: sleeping wired to instruments in an unfamiliar room alters sleep and alters dream content, an effect researchers document and try to correct for rather than eliminate. And what people report varies with how they are asked. None of this makes dream science unreliable. It makes it a field working at the edge of what is measurable, which is worth knowing when you read a confident headline about what dreams are for.

Frequently asked

What this is based on

• Standard polysomnography and laboratory awakening methods in sleep research
• The Hall and Van de Castle content-analysis coding system
• Francesca Siclari and colleagues' posterior hot zone research, and Yukiyasu Kamitani's group's fMRI dream decoding