Since the 1960s, the idea that our dreams take place mainly during REM sleep has dominated sleep science.
But neurobiologist Jean-Pol Tassin, director of research emeritus at INSERM, defends a radically different hypothesis: the dreams we remember would not be experienced in sleep, but manufactured at the time of micro-awakenings.
A scenario that may seem counter-intuitive... but gains in credibility once we examine its physiological underpinnings and explanatory power.
Editor's note: Very original and humane ideas, AI-assisted formatting.
The heart of the hypothesis: dreams are born on the edge of wakefulness
According to Jean-Pol Tassin, during REM sleep, two essential neuromodulators - norepinephrine and serotonin - cease activity altogether.
Result: no consciousness possible in this state.
Yet, to manufacture a dream that we remember, we need a brief window of consciousness: that of the micro-awakenings.
- A micro-awakening lasts a few seconds or less, often without you even realizing it.
- It is in this time that, according to him, the brain creates the dream - sometimes in less than a second - from a few images and available sensory or memory information.
This mechanism would also explain why some dreams seem perfectly integrated with reality.
Dreams often triggered by a real stimulus
Tassin cites numerous examples where a dream appears to have begun before the actual event... when it is in fact retroactively constructed after the event that woke us up.
Examples:
- Light on → dream of a dazzling spotlight.
- Sudden noise → dream of gunshot or door slamming.
- Drop of water on the face → dream of rain.
This instantaneous integration of real elements is much easier to explain if the dream is fabricated at the precise moment when the stimulus occurs, rather than several minutes earlier.
Good sleepers, bad sleepers: a key clue
Tassin observes that :
- For good sleepers (about 10-12 micro-awakenings per night) → few memorized dreams.
- For poor sleepers (around 17 micro-awakenings per night) → there are many more memorized dreams.
So it's the frequency of micro-awakenings, not the total duration of REM sleep, that would predict the amount of dreams remembered.
Content shaped by our latest neural networks
In this model, the brain draws on what Tassin calls "attractor basins": regularly activated sets of memories, images and emotions.
A micro-alarm activates one of them, producing a coherent dream... or a fanciful one.
Recent, significant or repetitive experiences increase the likelihood of their elements appearing in a dream.
Dominant model vs. Tassin model: the comparison
| Dominant REM model | Tassin model (micro-clocks) |
|---|---|
| Dreams, especially during REM sleep | Dreams made during micro-awakenings |
| Actual duration ≈ perceived duration | Actual duration < 2 seconds (perceived as long) |
| Integrated but secondary stimuli | Real-life stimuli that are often triggers |
| Dream frequency linked to amount of REM | Dream frequency linked to the number of micro-awakenings |
Stronger explanatory power
Compared to the dominant model, that of Tassin :
- Avoids the contradiction of a conscious dream in a supposedly unconscious state.
- Explains why we dream more during fragmented sleep (fever, discomfort, noise).
- Reflects the frequent integration of real stimuli.
- Justifies why dream content often reflects elements recently experienced.
- Predicts a very short real duration, with a posteriori reconstruction.
My personal experience: why the hypothesis speaks to me
As far as I'm concerned, I've always noticed a clear link between the quality of my sleep and the frequency of my dreams:
- Fragmented sleep (fever, discomfort, car journeys...) → many more dreams, often related to real elements experienced just before.
- Lack of sleep → heavy sleep, hardly any dream memories.
- Long, light nights → more memorized dreams.
These observations fit perfectly with Tassin's central observation: the more micro-awakenings, the more dreams are remembered.
Why there's no consensus (yet)
As Jean-Pol Tassin points out, the "REM = dream" link has been firmly established since the work of Michel Jouvet and features in most textbooks.
According to him, many researchers not working specifically on dreaming continue to teach it as an established fact, without revisiting the methodology that led to this conclusion.
It should be remembered, however, that this analysis reflects Tassin's position, and remains to be confirmed by wider historical and scientific examination.
Changing such an entrenched paradigm requires challenging established assumptions through a fresh and ambitious analysis, corroborated by solid new experimental evidence.
Tassin's hypothesis does offer an attractive explanatory framework, which deserves to be tested more widely. But for me, his theory, as I understand it, while very pertinent, doesn't totally bury the previous theory: the two can be reconciled.
My passionate hypothesis
For as long as I've been interested in the subject - about fifteen years - I've always imagined REM as a disorganized average of everything we've recently experienced.
A phase where the brain reactivates, mixes and reassembles our thoughts and experiences, randomly replaying the neural routes recently taken, often the most useful (but sometimes harmful), while gradually replacing those that have become obsolete, that are no longer activated.
And modern science is more along these lines, indicating, among other things, that the REM phase is most likely intended to consolidate memory and learning.
This mechanism, moreover, explains the popular (and for once accurate) idea that we retain while we sleep.
Where I agree with Tassin is that this intense intermingling has nothing to do with narrative: it's a subterranean work, invisible to consciousness. It's not exactly where we dream.
Where I disagree, however, is that this does not mean that REM plays no role in dreaming. The high level of brain activity during this phase cannot go unnoticed and play no role whatsoever; it persists during and for some time after waking up. In fact, when we've just woken up, it takes some time for us to actually... "wake up".
So I'll try to reconcile the two hypotheses on the basis of how I feel about the dream and how I conceptualize it.
My dream design scenario
For me, it is precisely during the irruption of a micro-awakening and particularly when it takes place right in the middle of REM, that the brain has the opportunity to consciousize what it was sorting out and consolidating unconsciously.
The dream would therefore not be not a continuous creation experienced during REM, but the conscious crystallization - in a few seconds -of a brain state in full reorganization.
And this REM brain state doesn't change instantly on awakening, it persists for a few moments (like retinal persistence or the so-called "ghosting" effect of a screen), because the activity was so vivid that some of the networks are still active for some time. This time lapse of up to a few seconds or even up to a few minutes allows replaying this scene of disorganized neuronal stimulation, but in a state of consciousness, which then tries as always to make sense of the whole. And of course, all this while incorporating the possibility of adding any immediate real elements that would have caused the narrative of this story to awaken.
And the whole forms a dream.
A story shaped from the reorganizing chaos of our brain, quite often, from recent facts, recurring ideas or deep-seated fears that our brain reinforces or reactivates during sleep.
This scenario also explains the unpleasant sensation of forgetting a dream on waking, as if the memory were slipping through our neurons. Depending on the sequence and timings involved, it seems plausible to me that the persistence of the REM neuronal state is not strong enough for consciousness to provide an interpretation. This could certainly be measured to put everyone in agreement. But the idea seems plausible to me insofar as a neuroscientist like Stanislas Dehaene defines us in a conference from 2015 consciousness as information capable of circulating in different brain areas, which therefore requires sufficient stimulation. In the absence of sufficient stimulation, we think we perceive something, but it slips our mind, we lose the idea.
My new definition of a dream
The Larousse definition would then be rendered invalid since dreams do not occur during sleep:
Dream male noun - 2. Psychic production occurring during sleep, and which can be partially memorized.
Dream - definition Larousse
Here's mine:
Dream male noun - Brief awakening into consciousness of disordered brain activity, which the mind immediately attempts to organize into a narrative.
Dream - Robin Labadie definition - August 12, 2025
What about you?
Have you ever noticed that :
- Do you dream more when you're ill?
- Do you dream more when your night is interrupted by wake-up calls?
- Do your dreams include things that happened just before you woke up?
- Do your dreams resemble your worries or your previous day?
If so, your experience may well support this hypothesis. And if not, that would be interesting to know too.
Share your observations in comments: they could enrich the discussion and, who knows, help test this model.
Conclusion: a model to follow closely
It's always been said that dreams are poorly understood... and perhaps for good reason: we haven't really challenged the dominant model for over 60 years.
One possible cause is that dreams no longer really interest researchers, while the general public, of which I'm one, remains fascinated by this nocturnal mystery.
Whether or not we agree with Tassin's vision, it raises a question that is as fundamental as it is fascinating:
What if our dreams weren't born in the night, but at the dawn of each awakening?
While we wait for science to settle the matter, this model has the merit of reopening the debate and offering a reading grid that speaks to physiology, logic... and the daily experience of many sleepers.
Further reading - Interview with Jean-Pol Tassin on Le Manal Show having inspired this article.
Leave a Reply