Why
dream-on ?

C’est probablement la question la plus ancienne qu’on ait posée à la nuit. Et la réponse honnête, en 2026, tient en une phrase : nous ne savons pas. Ce qui n’est pas du tout la même chose que de dire qu’on ignore tout. On sait où le rêve se loge, on sait de quoi il est fait, on sait ce qui le déclenche, et l’on dispose de plusieurs hypothèses de fonction, chacune appuyée sur des observations réelles.

What follows doesn't settle it. That's deliberate. An article claiming 'here's why you dream' would be lying to you. But there's a lot to tell, and some of it is frankly counterintuitive.

i.The dominant model: the dream, child of REM sleep

Since the work of the neurobiologist Michel Jouvet (1925-2017), who discovered and characterized the REM sleep in the late 1950s, the common view is simple: dreaming occurs during this phase, where the brain activates almost as if awake while the body is paralyzed by muscle atonia.

The argument is strong. Wake someone up during the paradoxical phase, and in the vast majority of cases they'll tell you a dream. Jouvet's experiment on the cat deprived of its atonia, which gets up while asleep to mimic hunting scenes, strongly suggests a scenario is playing out during this phase. And dream time seems to correspond to real time: a ten-minute dream takes ten minutes.

This model remains the one defended by most laboratories. Yet it has a serious challenger.

ii.The micro-awakening hypothesis

Jean-Paul Tassin, neurobiologist and emeritus research director at Inserm, a specialist in neuromodulators, proposes something else. His objection is logical before being experimental: during REM sleep, noradrenaline and serotonin neurons are shut down, so there is no consciousness. Yet a dream, since you can tell it, is a conscious phenomenon. Therefore, it cannot take place there.

According to him, the dream would be manufactured at the moment of micro-awakening, these brief ascents towards wakefulness that the brain imposes on itself about ten times a night, when neuromodulators abruptly switch back on.

It must be said clearly: this is a minority research hypothesis that contradicts the dominant model. We present it because it is published, argued, and fruitful, not because it is settled. Three of its consequences are worth exploring.

iii.Analog and cognitive: why we don't read in dreams

Still according to Tassin, the unconscious would process information in analogique (images, associations, without language), while the conscious state operates in cognitif (words, sequences). The dream would be the passage from one to the other, and this passage would be one-way.

Hence a verifiable prediction: you cannot read in dreams. Lire suppose de retenir un mot, puis d’y revenir, donc un aller-retour entre les deux modes. On peut apercevoir un mot isolé, jamais tenir une phrase entière, et les rares personnes qui y parviennent travaillent professionnellement le texte. Isabelle Arnulf, de son côté, relève le même fait depuis les banques de rêves : lire, compter et écrire sont ce que l’on fait le plus mal en songe, alors que ce sont des gestes quotidiens.

For those interested in lucid dream, it's a gift: trying to reread the same line of text twice is one of the most reliable reality checks there is. The text changes, blurs, or refuses to stay in place.

iv.What we dream of, in real life

Isabelle Arnulf, neurologist, heads the sleep disorders department at Pitié-Salpêtrière Hospital in Paris. Her field is the 'dream banks' compiled since the 1950s, thousands of narratives analyzed without interpretative bias. The portrait that emerges looks quite different from what one imagines.

• Dreams are primarily visual. Millions of fibers connect the eye to the brain, compared to about 80,000 for the ear, and far fewer still for smell. We therefore very rarely dream of smells.
• Dreams are ordinary. Conversations, commutes, work: the majority of dream content is mundane. Spectacular dreams exist, but they are the ones we remember and tell, hence a massive selection bias in our memories.
• They lean more towards the negative. Difficult interactions dominate pleasant ones there.
• Sexuality is rare in them. About 4% of dreams reported by men, around 0.5% by women. Counterintuitive, given what the Freudian tradition led one to expect.
• The color follows the times. Banks from the 1950s contain many dreams in black and white, mirroring the media of the time; they have almost disappeared. The environment imprints on dreams, which suggests today's screens do too.

Two observations go further. The first concerns the gaps between men and women, which remained stable over sixty years : les femmes rêvent d’autant d’hommes que de femmes, les hommes rêvent surtout d’hommes ; les décors sont plutôt extérieurs chez les uns, intérieurs chez les autres ; l’agressivité onirique est plutôt physique d’un côté, plutôt verbale de l’autre. Que ces écarts n’aient pas bougé alors que la place des femmes dans la société a profondément changé rend l’explication purement sociale insuffisante. Personne ne sait dire pourquoi.

The second is staggering. Paraplegic people walk, dance, and run in their dreams, including those that never worked. One proposed theory involves mirror neurons: we would spend part of our nights putting ourselves in others' shoes, and the dreamed body would be as much an observed body as a lived-in one.

v.Six possible functions, none proven

Here is the status of the hypotheses. None have reached a consensus, and nothing prevents several from being true at the same time.

• Consolidate the memory. Sleep clearly improves what has just been learned. Deep slow-wave sleep would consolidate knowledge, while REM sleep would consolidate motor skills. This is the basis of the old advice to study just before sleeping.
• Regulate emotions. Replaying a charged scene in a state where the body no longer reacts (in REM sleep, a dream fright doesn't accelerate the heart) could help digest its emotional load. "Sleep on it" might therefore have a physiological basis.
• Simulate the threat. Hypothesis of the Finnish researcher Antti Revonsuo : the dream would be a training simulator, where you rehearse responses to danger without risk. Consistent with the predominantly negative tone of dreams, and with the dreams of a child in danger so frequent among new parents.
• Keep sleeping. Tassin's proposition: dreaming would cushion the abrupt transition from the unconscious to the conscious during the micro-awakening, allowing you to fall back asleep. If the stimulus is too strong, the dream fails and you wake up for good.
• Nurturing empathy. It's the path opened by dreams of walking in paraplegic people: putting yourself in another's place, night after night.
• Serve creation. The dream state freely associates what waking life keeps separate. Many artists and researchers report finding an image or a solution there, these accounts being, by nature, hard to verify.

vi.Why these dreams, and not others

An image proposed by Tassin helps understand recurring dreams: memories would be stored like pools of varying depth. The more significant an event was, the deeper the pool, the more likely it is to be reactivated night after night. This explains the intensification of repetitive dreams after a shock.

Another elegance of this model: during micro-awakenings, the sense of gravity doesn't always reactivate at the same time as consciousness. Yet, without gravity, only two possibilities remain. You fall, or you fly. That might be why the chute and the vol are the two most universal dreams in the world, and why they are so often linked in the same story.

vii.So, why do we dream?

We don't know, and that's good news. It means the question is alive, that labs are working on it, and that what you note upon waking has more value than any dream dictionary. The dream probably doesn't have une function, like an organ would have a role: it's more like a very rich byproduct of the brain's nocturnal work, for which several uses have proven useful.

In the meantime, the only truly scientific thing you can do with your nights is to observe them. Note, date, reread after three months. Patterns emerge, and they are yours.