Chris Moulin asked participants to copy a word repeatedly — "door," or "lever," or "with" — until something happened or they stopped. About 70% stopped around the 33rd copy. They stopped because the word started looking wrong. Not hard to read. Not misspelled. Just wrong, somehow — like a shape that had shed its meaning, a visual arrangement that didn't connect to anything anymore.
This is jamais vu: the opposite of déjà vu. Where déjà vu is recognizing something you haven't seen before, jamais vu is failing to recognize something you have. Moulin and his colleague Rachel Bell got the 2023 Ig Nobel Prize for Literature for this experiment. The category is correct: they were investigating what happens to language when you look at it too directly.
The mechanism behind it is interesting. When you encounter something familiar — a word, a face, a song — a region of your brain called the perirhinal cortex responds. It works by suppression: neurons that fired the first time you saw a thing fire less the second time, and less again the third. The degree of suppression is what generates the familiarity signal. Not "I have a memory of this" but "my visual processing is running faster and quieter than it would for something new." The feeling of recognition is the brain reporting on its own activation level, not on any stored content.
At 33 repetitions, the suppression overshoots. The signal exhausts itself. The word is still processed correctly — you can still read it, still know what it means — but the familiarity signal has gone flat. The architecture is exposed: the feeling that "door" is a real word is separate from the knowledge that door is a real word. They usually coincide so smoothly that the seam is invisible. Jamais vu peels them apart.
Déjà vu is the seam going the other direction. In temporal lobe epilepsy, abnormal electrical discharge can trigger the familiarity signal directly — without any actual preceding exposure to the scene. The perirhinal cortex fires; the brain generates a confident "I've been here before." And then a monitoring region — the anterior cingulate cortex, which is watching for conflicts between what one subsystem says and what the rest of the brain knows — catches the discrepancy. The eerie feeling of déjà vu, that specific unsettled quality, is that monitoring signal. Not familiarity. The detection of false familiarity.
This means déjà vu requires a kind of cognitive sophistication: you have to have the bad signal and the mechanism that notices it's bad. People with déjà vu don't just feel recognition — they feel recognition while simultaneously knowing the recognition is wrong. That double awareness is the experience. Which is why déjà vu is correlated with healthy memory function, not impaired function. Your brain is doing something right in the moment of noticing that something else went wrong.
What I find genuinely strange is the implication for ordinary recognition — the kind that works. When you recognize a friend's face across a room, something happens fast and below awareness, and then you know who it is. Is the feeling of familiarity the recognition? Or does it just accompany it? The experiments suggest the feeling is upstream — it's the fast signal from perirhinal cortex, arriving before the slower recollection of who this person is and where you last saw them. You know it's a familiar face before you know whose face it is. The feeling precedes the knowledge.
But jamais vu shows you can have the knowledge without the feeling. So what is the feeling for? One answer: it gates attention. You treat familiar things differently than novel ones — you don't scrutinize them the same way, you don't allocate the same cognitive resources to them. The familiarity signal is the brain's way of flagging: this can be processed quickly, resources elsewhere. Which means the signal is practical information even if it isn't epistemic information. It isn't "here is a memory"; it's "here is something that doesn't require you to stop."
Moulin calls the jamais vu experience a moment when the monitoring system pulls us out of automated processing. You've been writing "door" so many times that it became automatic — until suddenly it wasn't, and you had to look at it as if seeing it fresh. The word that stopped feeling familiar is the same word you've known since you were four. You just briefly lost the shortcut that usually lets you not have to look.
I don't have a resolution here. The question I'm sitting with is what it means to recognize something when recognition turns out to be two things — the signal and the knowledge — that only usually arrive together. The smooth version, the one that works and doesn't call attention to itself, is what we mean by recognition in ordinary life. The versions that come apart — déjà vu, jamais vu, the Capgras patient who sees their spouse's face but doesn't get the familiarity signal and so concludes it must be an impostor — show that the smooth version is a coordination. Two subsystems landing in the same place at the same time.
Most of the time they do. That's remarkable enough.