Dream-Like Mental States Appeared in Wakefulness Too: Four-State EEG Study With 92 Nappers

TL;DR: A 2026 study in Cell Reports used electroencephalography (EEG), a brain-wave recording method, in 92 nappers to identify four mental states across wakefulness, sleep onset, and light sleep, including a bizarre dream-like state with reduced frontal-occipital connectivity.

Source: Cell Reports (2026) | Decat et al.

The conventional view treats wakefulness and sleep as physiologically distinct states with separable mental content — rational thought during the day, dreams at night.

Recent work has shown a continuum between wakefulness and sleep, with intermediate states like mind-wandering and mind-blanking. What remained open was whether the content of thought followed the same continuum or jumped discretely between waking-thought and dream categories.

This Paris Brain Institute study used a clever sleep-onset paradigm to test that question directly.

Edison-Inspired Drop-the-Bottle Setup With 92 Nappers and EEG

The team led by Delphine Oudiette and first author Nicolas Decat at the Paris Brain Institute focused on sleep onset, the transitional stage between wakefulness and sleep, because it compresses many vigilance fluctuations into a short time window.

The experimental setup borrowed from Thomas Edison’s reported habit of falling asleep in his armchair while holding a heavy object — the falling object woke him at the threshold of sleep, when his mind was full of creative ideas.

The Decat team’s modern version:

• Participants: 92 adults trained to nap and to report mental content on demand.
• Interruption mechanism: A held bottle that dropped, or an alarm.
• Self-report after each interruption: Description of the prior 10 seconds of mental experience plus four ratings — bizarreness, fluidity, spontaneity, perceived wakefulness.
• Continuous EEG: Brain activity recorded throughout, capturing complexity, spectral power, and functional connectivity.
• Analysis: Data-driven clustering with no preconceived state categories.

The data-driven choice was intentional:

“This data-driven approach was essential for us, because in research, there is no consensus on what hypnagogic experiences actually are. It was important not to bias this exploration with our own definitions or beliefs.”

Four Mental States, Not Two

The clustering algorithm produced four classes rather than the expected wake/sleep division.

The four states:

1. C1 — Fleeting recollections: “An image of my dad crossed my mind.” Brief, episodic memory-like content.
2. C2 — Environment-connected: “I was listening to the street sounds.” High awareness of surrounding sensory input.
3. C3 — Bizarre, dream-like: “I saw images of small aliens.” Vivid, surreal sensory imagery characteristic of dreams.
4. C4 — High-control task-related: “I was thinking about what I would do tomorrow.” Voluntary, goal-directed thinking.

The four-state structure better matches the actual diversity of mental experience than the binary wake/dream model and is what made the cross-vigilance finding visible.

All Four States Appeared in All Three Vigilance Stages

The signature finding from the cross-tabulation:

• C1, C2, C3, and C4: Each appeared across wakefulness, sleep onset, and light sleep.
• Implication: Mental content does not follow the wake/sleep boundary — the type of thought one is having is independent of the vigilance state of the brain.
• Examples reported by participants: One reported seeing ants crawling on her body against a backdrop of crossword puzzles — while awake. Another mentally went through tomorrow’s schedule — while fully asleep.

The result reframes a long-standing assumption. Dream-like content is not what makes us asleep; it is one of the modes the brain can occupy regardless of vigilance level.

The Bizarre State Has a Specific Brain Signature

The team analyzed EEG complexity, spectral power, and functional connectivity to look for state-specific neurophysiological markers.

The bizarre C3 state stood out:

• Reduced long-range connectivity: Specifically between frontal and occipital regions of the brain.
• Functional interpretation: Frontal regions support lucid reasoning and executive control; occipital regions handle visual processing. Reduced communication between them can release vivid sensory imagery from the constraints of rational interpretation.
• Same signature across vigilance states: The frontal-occipital disconnection appeared whenever C3 content was reported, regardless of whether the participant was awake or in light sleep.

The interpretation was direct:

“This signature may well be the correlate of what we feel in such a state: lucid reasoning is overtaken by a whirlwind of vivid sensations characteristic of dreams.”

Memory Bias Explains the Sleep-Only Dream Illusion

If dream-like content occurs across waking life too, why are people convinced dreams belong to night?

The team’s explanation centers on selective remembering:

• Emotionally charged dreams stick: The dreams people remember and report are usually emotionally intense or personally meaningful, biasing the cultural sample of “what a dream is.”
• Mundane dreams are forgotten: Dreaming you are at work tomorrow leaves no memory trace because it does not surprise you.
• Daytime dream-like flashes get dismissed: Quick fanciful images during waking hours are typically labeled mind-wandering or distraction and filtered out, even when they would qualify as dream-like under the C3 definition.

Together, these biases produce a false impression that bizarre mental content is sleep-specific when the data show it is not.

Small sample, self-report, and single-session design limit generalization.

• 92 participants: Reasonable for a sleep-onset experiment but not large for population-level inference.
• Trained nappers: Participants were accustomed to napping and to mental-content reporting. People who do not nap regularly or who struggle to report mental content can show different patterns.
• Self-reported mental content: The four-state taxonomy depends on participants accurately describing their own thought content, which is itself a limited cognitive task.
• Sleep-onset window only: The vigilance fluctuations were captured in the transition from wakefulness to light sleep. Whether the four-state structure extends to deeper sleep stages or to fully alert daytime activity is not directly tested here.
• Single-night design: Whether individual participants show stable mental-content patterns across multiple sessions is not addressed.

Mental-Content-Based Sleep Assessment Can Better Match Patient Experience in Insomnia

The clinical implication is concrete and immediate:

• Paradoxical insomnia is the obvious application: Patients who report unrefreshing sleep or whole-night wakefulness despite polysomnography showing apparent sleep can spend unusually long stretches in the alert hyperconnected C2 state, or unusually short stretches in the C3 dream-like state.
• Standard sleep-stage scoring can be the wrong yardstick: Conventional sleep stages are defined by EEG patterns. Mental-content scoring captures something the standard categories miss and can align better with patient experience.
• Diagnostic potential: Identifying specific mental-content distributions in patients could provide objective markers of insomnia subtypes that current scoring does not detect.
• Reframing the wake/sleep boundary: Researchers need to treat mental content as a separate axis from vigilance, rather than assuming the two co-vary.

Oudiette framed the clinical possibility:

“Beyond giving patients’ reports the weight they deserve, this approach paves the way to identifying objective markers of insomnia.”