Draft — This post is a work in progress.
Some Thoughts on Déjà Vu
Summary
-
What it feels like: a spike of familiarity with no retrievable episode.
-
Best current model: a memory-monitoring mismatch — an errant familiarity signal (medial temporal system) that your frontal monitoring system flags as inappropriate.
-
Mechanisms aren't single-cause: evidence spans timing desynchrony, network misfiring (esp. temporal lobe), implicit/gestalt familiarity without recollection, and attentional/novelty effects.
-
Health note: common, usually benign; frequent/intense déjà vu can appear in temporal-lobe epilepsy or with some dopaminergic/serotonergic drugs.
I found myself feeling déjà vu for the first time in years. It was quite common in my middle school and early high school days, but now it had become rare. So I got curious and asked GPT-5: "hey, what is known about déjà vus?"
The first answer was the short-term → long-term memory "shortcut." It feels right... but I couldn't find peer-reviewed research that endorses that as the leading mechanism for everyday déjà vu. It appears in popular explainers, not in modern reviews.
What this says is that there might be a pathway straight to long-term memory. The memory is stored there, and when our familiarity sense gets a perfect match, we get our déjà vu.
Honestly, for me this feels absolutely right.
Only a couple of questions: then why do we feel so much discomfort, and are we so certain that there is no way we lived that before?
We could answer this with: well, maybe the brain doesn't find memories conducive to this one and that triggers discomfort? I think it tracks.
But then I did more digging, and I found the more accepted answers were given here:
(A) Memory/Gestalt (implicit familiarity) + a monitor. Your brain stores lots of layouts/relationships. When a new scene's structure rhymes with prior ones, your familiarity system pings "known," but no specific episode is found. That mismatch—familiarity without a source—is déjà vu. Lab VR work shows layout similarity alone boosts déjà-vu reports without successful recall.
(B) Dual-processing (timing desynchrony). A fast familiarity route can briefly outrun slower, detailed recollection. For a second you get "I know this... from where?"—until recollection catches up or your brain's monitor calls it a false alarm. Recent reviews frame déjà vu as this kind of conflict between a familiarity signal and a concurrent sense of novelty.
(C) Neurological (misfiring networks, clinical edge cases). In temporal-lobe epilepsy (and sometimes with certain drugs) memory circuits can over-generate familiarity; electrical stimulation of medial-temporal regions can evoke déjà-like states. Helpful to triangulate circuitry—but a different beast from the occasional, healthy experience.
(D) Attentional / novelty & expectation. Expected/patterned inputs bias the system toward familiarity; distinctive/unexpected inputs bias recollection. This explains why tidy, template-like spaces often spark déjà vu more than one-off, idiosyncratic scenes.
Do background and lifestyle matter?
Survey and individual-differences work suggest people who travel more report more déjà vu (more near-matches of scenes). Classic reviews also note a positive relationship with education/socioeconomic level—which may reflect more exposure, better noticing/reporting, or both.
Quick FAQ
Is the "short-term → long-term shortcut" real? Not as a peer-reviewed, leading explanation. It's a popular metaphor; the research picture favors pattern-driven familiarity + monitoring conflict.
Why does it feel so certain? Because the familiarity route is built for speed. When it fires hard on a pattern match, it feels like you "just saved this to memory," even though no episode is retrieved.
Why more as a teen? More novel environments and more excitable circuits; reports decline with age.
Graphics
Figure 1 — My initial (now rejected) "two-memories shortcut" idea
A 2-lane diagram: Short-term buffer → Long-term store, with a dashed "shortcut" arrow labeled "popular hypothesis (weak evidence)." Catchy metaphor, not supported by peer-reviewed mechanisms for everyday déjà vu.
Figure 2 — Gestalt + monitor (preferred everyday model)
Left: Generator (medial-temporal familiarity) lights up on pattern/layout match. Right: Monitor (prefrontal) queries source memory → "no episode" → mismatch felt (déjà vu). Includes a tiny room-layout pair showing identical geometry (sofa–lamp–art triangle) with different objects.
Figure 3 — Peer-reviewed schematic
A simplified re-draw based on review figures: medial-temporal nodes (perirhinal/entorhinal/anterior parahippocampal) feeding a prefrontal monitoring node, annotated "errant familiarity → interpreted as inappropriate."