You're at a family dinner, and you tell the story. You know the one — the story you've told forty times, the one with the timing you've got down, the one where your brother always interrupts at the same place.

Except tonight he doesn't interrupt. Tonight he says, quite calmly, "That's not what happened. I was there. That's not how it went."

And something drops in your chest, because you can see it. You can see the kitchen. You know what the light was like.

You are not lying. He is probably not lying either. And neither person's certainty can settle the details by itself.

Here's the part that should bother you more. If memory were a recording — a video file sitting on a hard drive somewhere behind your eyes — then a story told forty times would be protected by repetition. That's how recordings work. Playing a file doesn't change it.

Unless memory was never a recording to begin with.

Memory is reconstructive, not literal playback. The archive metaphor is useful if we add one important limit: under some conditions, retrieving a memory can make parts of it open to updating before it stabilizes again. Not every retrieval does this, and not every detail changes. But the act of remembering can interact with what happens after the original event.

Let's go into the archive.

The book comes off the shelf

For most of the last century, the assumption was that memories get made and then set, like concrete. The technical word is consolidation: a memory forms, it's fragile for a while, and then it hardens into something stable and finished. After that it just sits there, and remembering is a matter of going and having a look at it. Worth naming the shelf before we take the book off it. Filing a new episode — the kitchen, the light, your brother's face — runs through the hippocampus, a curled structure buried in each temporal lobe and named, a little generously, after the seahorse someone thought it resembled. Damage the hippocampus and the ability to form new conscious memories falls apart, while memories from the distant past are often relatively spared.8 The standard account of why is systems consolidation — a slow, years-long reorganization, a different process from the synapse-level hardening described above: the hippocampus doesn't keep the book for good. It guides material toward the neocortex, the wrinkled outer bulk of the brain that was actually involved from the moment of learning too, until, in the strongest version of the theory, a memory no longer needs the hippocampus to be pulled up.8 How complete that handoff ever gets is genuinely unsettled: a rival account argues that memories holding onto their vivid, specific detail stay hippocampus-dependent for as long as they last, and only the smoothed-down gist version moves out — a disagreement the field has been having for decades.9

The finding that broke that picture is more conditional than the version that usually travels online: reactivation can, under the right conditions, make a consolidated memory temporarily open to disruption or updating.

In many animal experiments, a reactivated memory enters a time-limited state in which an intervention can strengthen, weaken, or alter its later expression. Restabilization is called reconsolidation. But retrieval is not an automatic unbind button — memory age, strength, prediction error, the kind of memory, and the reactivation procedure can all matter.23

That doesn't mean the stable memory disappears every time you remember it. Human studies show many findings consistent with reconsolidation, but inconsistent results and alternative explanations mean researchers cannot infer one universal neurobiological process from behaviour alone.5

Human memory is still plainly vulnerable to information arriving after an event.7 In misinformation experiments, repeated misleading details become more familiar, more likely to be reported, and more confidently endorsed.6 That doesn't prove telling a family story forty times steadily erases the original. It does show why repetition is not independent confirmation — and why two honest people can leave the same kitchen with different reconstructions.

What the rebinding costs

So what can the brain be doing during that window? One animal study gives an unusually detailed answer.

In mice trained on inhibitory avoidance, an eLife study found two processes happening at the same time after retrieval, under conditions that strengthened the memory.1

The first is protein degradation — the brain actively breaking down existing proteins at the synapse. Taking things apart. Unstitching the spine.

The second, simultaneously, is switching on genes to build new proteins. New material. Sewing it back together.

Demolition and construction, running at once, in that experimental memory system. That is not what playing a file looks like. It's closer to renovation.

In that study, reactivating the memory under the right conditions didn't just preserve its later expression — it strengthened it through the same tear-down-and-rebuild process. The work was also divided across regions: the amygdala was required for reconsolidation and enhancement, while the hippocampus and medial prefrontal cortex were required for enhancement but not the basic reconsolidation effect. In that one paradigm, restabilizing and strengthening were not identical jobs.

Which brings me to an argument I want to challenge: I've gone over it so many times, I know it's right.

Repetition can strengthen access and familiarity. It can also repeat an error, or fold in information that arrived later. Experiments on repeated misinformation show repetition increasing both susceptibility and confidence. Rehearsal is not proof of preservation.

That doesn't make your most rehearsed memory your least accurate. And confidence does carry useful information in some settings, especially when recorded early and without contaminating feedback. It simply isn't proof that every detail survived unchanged.

What we actually know, and about whom

Now I have to be honest with you, because this is the part where a lot of science communication quietly stops talking.

The study I've been describing — the protein degradation, the gene expression, the division of labour between amygdala and hippocampus — was done in rodents. And not on the memory of a family dinner. On a very specific, very controlled kind of aversive learning, in a laboratory, in an animal whose whole brain would fit inside a walnut.

That's not a reason to dismiss it. Animal studies let researchers look directly at synapses and interrupt molecular processes in ways that would be unethical in people, and the mechanisms are well described in some animal paradigms.

But there's a genuine distance between a reactivated aversive memory in a rodent became labile under this procedure and your recollection of your grandmother's kitchen is molecularly rewritten every Christmas. Human behavioural work establishes reconstructive memory and susceptibility to post-event information. It does not let us assign the mouse mechanism to each autobiographical recollection.

And the field itself hasn't settled. There's an active argument in the literature about whether what looks like a memory being rewritten is really the old trace being modified, or a new competing trace being laid down alongside it and winning the race to be retrieved.4 Those two explanations predict very similar behaviour and are extremely hard to tell apart from the outside.

I'm telling you this not to undercut the piece. I'm telling you because your memory is unreliable has become the kind of line people repeat with enormous confidence — which is a slightly funny way to hold a belief about the unreliability of confidence.

Back to the table

So where does this leave you, sitting at that dinner, with your brother telling you the kitchen was a different colour?

It leaves you with a memory system that reconstructs rather than replays. That's normal — not evidence that either of you is lying, or that either brain is failing. But resist the equally simple myth that every act of remembering rewrites everything.

Hold certainty with enough humility to ask what came from the event, what came later, and whether repetition has made a detail familiar rather than independently verified. When someone who was also there remembers it differently, disagreement alone does not reveal who is wrong, or why.

And don't turn reconsolidation into a do-it-yourself promise that revisiting a painful memory in the right mood will rewrite it safely. Human clinical findings are mixed, boundary conditions matter, and memory-focused treatment belongs with a qualified clinician.

You cannot go back and change what happened. Remembering is not literal playback, and it is not unlimited editing either.

The useful question is smaller and more honest: what do I remember, how do I know, and what may have entered the story since?

Sources

  1. Enhancement of fear memory by retrieval through reconsolidation. eLife (2014).
  2. Memory retrieval and the passage of time: from reconsolidation and strengthening to extinction. Journal of Neuroscience (2011).
  3. The timing of multiple retrieval events can alter GluR1 phosphorylation. Learning & Memory (2012).
  4. Reconsolidation versus retrieval competition: rival hypotheses. Behavioral and Brain Sciences (2015).
  5. Human memory reconsolidation: a guiding framework and critical review of the evidence. Psychological Bulletin (2018).
  6. Repetition, not number of sources, increases both susceptibility to misinformation and confidence in the accuracy of eyewitnesses. Acta Psychologica (2012).
  7. Planting misinformation in the human mind: a 30-year investigation of the malleability of memory. Learning & Memory (2005).
  8. Squire LR, Genzel L, Wixted JT, Morris RG. Memory consolidation. Cold Spring Harbor Perspectives in Biology (2015).
  9. Moscovitch M, Cabeza R, Winocur G, Nadel L. Episodic memory and beyond: the hippocampus and neocortex in transformation. Annual Review of Psychology (2016).

General education about how minds work — not medical advice, and not an assessment of you.