Dreams in REM and non-REM sleep

What a sleep stage describes

NHLBI distinguishes REM from three non-REM stages using recorded brain activity and eye movements. NREM stage 1 marks the transition into sleep; stage 3 is slow-wave sleep. Cycles change across the night, with more deep sleep earlier and more REM later. These are measured sleep categories, rather than labels assigned from a dream’s plot.

Direct evidence beyond REM

Siclari and colleagues compared reports after awakenings with high-density EEG. Reports of experience, in both REM and NREM, were associated with reduced low-frequency activity in posterior cortical regions. Monitoring that activity also helped predict subsequent NREM reports under laboratory conditions.

This establishes that research on dream experience cannot simply substitute “REM” for “dreaming.” It does not produce a recording of the entire dream, identify its personal meaning, or show which brain activity is sufficient to explain every aspect of experience. The published comparisons also use different participant subsets, so each analysis has its own denominator.

An earlier comparison makes report length visible

Cavallero and colleagues' 1992 abstract describes a laboratory comparison in which volunteers were awakened from REM and slow-wave sleep on separate nights, with the order counterbalanced. Judges rated the resulting reports, and participants identified possible memory sources. REM reports were longer on average in that particular comparison.

The complete paper was unavailable, so the historical study is retained as limited context rather than a fully inspected classification test. Counterbalancing makes the collection procedure relevant, but an average length difference does not identify the stage behind one person's account. Brief imagery does not establish non-REM sleep, and a long narrative does not prove REM. The paper's historical slow-wave terminology also should not be read as a new set of current sleep stages.

Three kinds of answer after waking

The 2025 DREAM database standardizes reports as experience, experience without recalled content, or no experience, allowing combinations when records are ambiguous. Its broad definition includes thought-like and fragmentary experiences. The original datasets use differing procedures, so harmonized labels improve comparison without eliminating every collection difference.

An illustration: the same sentence, different missing information

Consider the invented note “I was on a blue train.” It contains a setting and a colour. It does not tell us when the scene occurred, whether it felt long, how much was forgotten, or what a physiological recording would show. Adding “it was incredibly vivid” makes the description richer, but does not supply the missing measurement.

Now consider “I know something was happening, but I cannot remember it.” That is a different kind of report from the train sentence. It is also different from confidently describing an empty interval. For personal notes, the practical option is to keep the wording that fits your memory. You do not have to reconstruct a story in order to give the experience a respectable label.

What an average can and cannot classify

Here is an original statistical illustration unrelated to measured dream data. Suppose group A writes accounts averaging 100 words and group B averages 60. There could still be many 40-word accounts in A and many 120-word accounts in B. Knowing the two averages is not enough to classify a new 80-word account. You would also need the spread of results, overlap and a tested decision rule.

Apply that reading question whenever a headline contrasts “REM dreams” with “NREM dreams.” Look for the actual feature measured: report length, a rating, an image category or an electrical signal. Then look for the evidence that the feature classifies an individual. A descriptive difference between samples should not quietly become a test offered to readers.

A clock or graph needs its own evidence

For an optional record, write down when you noticed waking without calculating backwards to invent the dream’s beginning. “I remembered it at seven” is an observation about recall time; it does not establish when every part of the scene happened.

Consider an invented example: at seven you remember the blue train, while a tracker labels the preceding interval as light sleep. The account records remembered content and the graph records an estimate. Neither identifies the exact moment of the scene, and vividness alone cannot settle a disagreement.

Ask how the device produced its labels and what they were validated against. Keep the remembered account and displayed estimate separate. The tracker guide explains how to read validation; the sleep-stages guide explains the physiological labels without turning a chart into a personal timetable.

REM rebound is a measured change, not a vividness label

In a small laboratory experiment, Endo and colleagues selectively interrupted REM sleep in eight healthy young men. During the first recovery night, REM amount reached about 140% of their baseline amount. That means roughly 40% more than baseline, not that REM occupied 140% of the night. The study compared physiological recordings, not a count of remembered stories.

This provides a concrete example of what researchers call REM rebound after suppression. It does not establish that an unusually vivid morning proves rebound, that every kind of sleep disruption has the same effect, or that lost dreams are being replayed. A self-described increase in recall remains a different measurement.

Keep the comparison explicit: which recorded quantity, relative to which baseline, after which circumstances? Do not recreate sleep deprivation to test a dream theory. An optional note about a memorable scene needs no physiological label.

Choose the question that still matters

The related nap, vivid-dream, lucid-dream and dream-recall guides offer different routes for further reading. Choose the question that interests you without first assigning your own sleep stage. You can read about a remembered experience on its own terms.

A short note can contain an image, a feeling and uncertainty about timing. There is no requirement to install a monitor, wake repeatedly or compare every morning with a stage chart. If your question is already answered, you can stop reading without turning an ordinary rest into a measurement project.

Sources & their limits

These sources support the statements indicated above. Their authors do not endorse this website or provide an assessment of your experience.

1. NHLBI — How Sleep Works: Sleep Phases and Stages

   Official definitions and changing stage distribution.

   Cited in this guide · 1 place

   • What a sleep stage describes
2. Siclari et al. (2017) — The neural correlates of dreaming

   Reports and EEG correlates across REM and NREM.

   Cited in this guide · 2 places

   • Direct evidence beyond REM
   • Can dreaming occur outside REM sleep?
3. Wong et al. (2025) — A dream EEG and mentation database

   Report classifications and differing contributing protocols.

   Cited in this guide · 2 places

   • Three kinds of answer after waking
   • What does experience without recall mean?
4. Cavallero et al. (1992) — Slow wave sleep dreaming

   Original abstract only; counterbalanced laboratory REM/slow-wave report comparison. Complete paper unavailable; group-average report length, not an individual sleep-stage classifier.

   Cited in this guide · 2 places

   • An earlier comparison makes report length visible
   • Does a very long report prove REM sleep?
5. Endo et al. (1998) — Selective REM sleep deprivation in humans: effects on sleep and sleep EEG

   Original controlled deprivation experiment: Methods, Table 1 and physiological recovery outcomes in eight healthy young men; no dream-content outcome.

   Cited in this guide · 1 place

   • REM rebound is a measured change, not a vividness label

Put the context beside an experience

Continue through the science library