What is déjà vu?

What counts as déjà vu?#

Two things together. Cleary and colleagues define it as “the striking sense that the present situation feels familiar, alongside the realization that it has to be new”. Aitken and colleagues put it as a conflict between a sense of familiarity and a concurrent sense of novelty. Illman and colleagues call that second part critical. A false feeling of familiarity with no sense that it’s false is an ordinary memory error, like picking “sofa” as a word you studied when it was “settee”.

The word comes from French, and Wikipedia credits its first use to the philosopher Émile Boirac in 1876. It has relatives, which are easy to mix up.

Familiarity and recollection are the split behind the remember/know test: knowing you’ve met someone against remembering where.

Why does déjà vu happen?#

No one has settled it. Brown’s 2003 review sorted the explanations into four groups: dual processing, neurological, memory and attentional. The work we read has mostly gone into the memory group and a conflict account that came later.

A real resemblance you can’t place

On this memory account, something about the scene does match your past, and the matching experience doesn’t come to mind. Cleary and colleagues tested the idea in virtual reality in 2012. A new scene laid out like one people had viewed earlier raised familiarity and reports of déjà vu, when they couldn’t recall the earlier scene. The closer the match, the stronger the effect.

The brain noticing a conflict

This account focuses on the second half of the experience, knowing the feeling is wrong. Urquhart and colleagues (2021) gave 21 people a false-memory task that produced familiarity, along with cues that it was mistaken, and 16 of them reported déjà vu. In a scanner, the regions linked to that memory conflict included the anterior cingulate, medial prefrontal and parietal cortex. They describe it as the first experiment to image a lab version of déjà vu in healthy volunteers.

Aitken, Jentzsch and O’Connor (2023) propose that this could explain the age pattern: conflict monitoring and déjà vu both peak in young adulthood and then decline. Theirs is a narrative review, not a test of it.

A stray signal of familiarity

Here the feeling needn’t come from the scene. Illman and colleagues (2012) describe familiarity decoupled from what’s in front of you: “free-floating and unrelated to the sensed environment”, “a random top-down brain event”. They allow for the forgotten-experience case too, and call recognition-without-identification work, which includes Cleary’s scene studies, the nearest lab analogue. They point to the parahippocampal region, beside the hippocampus. Much of this evidence comes from epilepsy research. In Bartolomei and colleagues’ patients, who had electrodes placed for investigation, stimulating the rhinal cortices, next to the hippocampus, brought on déjà vu more readily than stimulating the hippocampus itself. How far that carries to everyday déjà vu isn’t established.

Two older ideas

Dual processing, in Brown’s words, is “2 cognitive processes momentarily out of synchrony”. Wikipedia traces one version to Robert Efron in the 1960s, who proposed that delayed signals get processed as two experiences, the second seeming to re-live the first. The attentional account is “unattended perception followed by attended perception”. We didn’t read tests of either.

These aren’t necessarily rivals, and that part is our reading: a resemblance or a stray signal could supply the familiarity, and a conflict check could be what makes it feel strange.

How common is it, and who gets it?#

• Brown’s review: about 60% of people have experienced it, and it gets less frequent with age.
• The same review links it to stress and fatigue, and to higher socioeconomic level and education.
• O’Connor and Moulin (2013), with 206 people, replicated a link with how often people travel. An old example they cite: 11% of people who didn’t travel in a 1951 study, against 32% of those who travelled up to five times a year.

How often is less well measured. A survey of Russian medical students, in a short conference abstract, describes it as “quite rarely (from 2 to 10 times over many years)” for most respondents, mainly in childhood and adolescence.

Most déjà vu is harmless, and it can also be a symptom of epilepsy. An American Academy of Neurology article, drawing on Jean Khoury, a Cleveland Clinic neurologist, says seizure-related déjà vu can last a minute or two and often comes with other symptoms, such as nausea, a rising feeling in the abdomen, repetitive hand or mouth movements, or losing awareness, and that a neurologist should check out any combination of these.

Why does it feel like you know what’s next?#

Many people report that during déjà vu they feel they know what’s about to happen. Cleary and Claxton tested it in 2018 with navigation paths that resembled earlier paths people couldn’t recall. The resemblance raised reports of déjà vu, and in that state people felt more sure of the next turn. They did no better than chance at predicting it.

McNeely-White and Cleary (2025) found the same with sound. Déjà entendu reports came with a stronger feeling of knowing the next note. In one experiment accuracy stayed at chance, and in the other the next note’s location was random, so nothing could have predicted it.

What isn’t known?#

• How far to trust the reports. Everything above rests on what people say they felt. Jersakova and colleagues ran four experiments with an ordinary recognition task not expected to bring on déjà vu or tip-of-the-tongue states. Between 32% and 58% of people still reported déjà vu or tip-of-the-tongue, and in the first experiment 41.1% of the control group reported déjà vu. In a later experiment, people told to flag it whenever it happened were less likely to report déjà vu afterwards than a group asked every 12 trials. Those groups came from separate experiments, and tip-of-the-tongue reports showed no significant difference.
• Whether the lab versions are the real thing. Urquhart and colleagues call theirs an analogue, and it isn’t known how closely any lab method matches the experience on a street or in a hallway.
• Whether people who get it more have different memories. O’Connor and Moulin predicted a link with recollection or familiarity on a standard word test, and found none. They suggest a mismatch between memory signals and monitoring that such tests don’t pick up.
• Which brain regions matter. Illman and colleagues note inconsistent findings across the temporal lobe, and say part of that comes from poor classification of déjà experiences.
• How it relates to jamais vu. In Moulin’s study, people with more déjà vu also had more jamais vu in daily life, yet the lab version didn’t show déjà vu’s links to age and dissociative experiences. So how close the two are is still open.

Try it

We can’t give you déjà vu, so try its opposite#

We don’t have a tested way to give you déjà vu on a web page. The methods above use purpose-built scenes, paths, sounds and memory tasks, and they still rest on what people report. The reverse, jamais vu, is when something you know well turns strange, and in one study copying a word was enough to bring it on in some trials.

This follows Moulin and colleagues’ word task, which won a 2023 Ig Nobel prize in literature. We read only the abstract, so the words, the typing and the 60-copy limit are our choices. The authors describe what people reported as jamais vu, and their title also names word alienation and semantic satiation. The button is itself a prompt, and Jersakova’s results show that how people are asked can shape what they report, so treat your result as a curiosity. Your words stay on this page, and we only count that a demo was started.

What do people get wrong about it?#

Whether “already lived” counts as déjà vu

Wikipedia’s first sentence defines déjà vu as “feeling like one has lived through the present situation in the past”. In the split Illman and colleagues propose, a sense of having lived it is closer to déjà vécu, and déjà vu is false familiarity without that detail. They call their classification preliminary, and suggest déjà vu, déjà vécu and fragments of real memory probably lie along a continuum. It’s still worth asking which one a study measured.

That there’s one number for how many people get it

Brown gives about 60% and Wikipedia about two-thirds. The Russian student survey, one university’s students, went from 82% in 2001–2004 to 98.5% by 2015. Samples and wording differ across these, and Jersakova’s results show that the method of asking can move reports.

That the feeling of knowing what’s next is knowledge

The association with premonition is real as a feeling, and Cleary and Claxton’s abstract calls it a remaining puzzle. In the two sets of experiments above, the feeling didn’t come with any accuracy. Those are the only direct tests we found, on a limited set of tasks.

Do AI systems get déjà vu?#

Nothing we read suggests it. The closest parallel is a memory system that retrieves by similarity: a new input can closely match something stored, the way Cleary’s new scene matched an old one. That’s a score on a match, and there’s no evidence of a strange feeling attached to it.

Counterparts, the memory system whose site this is, keeps what the AI writes down in its own words and brings back what fits. The Field Guide follows mechanisms that AI systems do borrow, and the directory compares the systems.

Questions people ask#

Questions about how any of this carries over to AI? Ask in r/AIMemorySystems.