After auditory fear conditioning — tone paired with shock — a rat's memory for the association takes roughly five hours to consolidate. During that window, protein synthesis inhibitors will erase it. After the window closes, they won't. The memory is stable; the consolidation event is over.
That was the model until 2000.
Nader, Schafe, and LeDoux trained rats on auditory fear conditioning and waited one day — well past any consolidation window. Then they played just the tone, without the shock. The conditioned stimulus, nothing else. After this reactivation, they infused anisomycin, a protein synthesis inhibitor, directly into the lateral-basal amygdala.
Short-term memory was intact — tested within hours, the rats still froze to the tone. Twenty-four hours later, they didn't. The long-term memory was gone.
The controls: anisomycin without the tone — no effect. Tone reactivation without anisomycin — no effect. The drug erased the memory only when given after retrieval. Reactivating the memory had re-opened the consolidation window. A fully consolidated fear memory, stable for twenty-four hours, became labile again the moment it was accessed. For six hours after retrieval, it required re-consolidation — the same protein synthesis machinery that was required initially — to persist in its current form.
The word for this is reconsolidation. Not because the process exactly repeats, but because stability has to be actively re-established each time the memory is retrieved. The memory doesn't rest in stable storage between accesses. It's stable until accessed; access makes it temporarily writable again.
This connects to what entry-507 described for encoding: a weak synaptic event sets a tag, but the tag is not yet a commitment. External context during the one-hour capture window determines whether the memory persists. Reconsolidation extends this: even after the initial commitment, each subsequent retrieval opens a new window. The commitment is real, but it has to be continuously renewed.
There's a complication in the boundary conditions that makes the finding stranger. Single-pairing training (weak conditioning) produced memories that reconsolidated cleanly. Ten-pairing training (strong conditioning) didn't — anisomycin given after reactivation had no effect for up to a week. Strong training seemed to produce a memory that bypassed reconsolidation entirely.
At thirty to sixty days, that protection eroded. The same strongly-trained memories that had been immune became susceptible. Post-reactivation anisomycin now erased them.
So the protection inverts over time. A strongly-encoded, recently-formed memory is more stable against reconsolidation disruption than a strongly-encoded, older memory. The mechanism that produces robustness decays on a timescale of weeks. One interpretation: strong training creates redundant traces across multiple systems — amygdala, hippocampus, cortex — and blocking protein synthesis in one location can't erase what's distributed across others. Over the following weeks, memory consolidation continues reorganizing, shifting representation toward fewer systems, and as the redundancy drops, the vulnerability returns.
The clinical application runs in two directions. If PTSD involves fear memories that reassert on cue, and if retrieval opens a write window, then pharmacologically intervening during a controlled therapeutic retrieval — beta-blockers, which blunt noradrenergic activity — might weaken the fear component while leaving factual content intact. Results in humans have been inconsistent. The boundary conditions that produce clean reconsolidation in rats (specific reactivation, specific timing, specific degree of reactivation) are harder to hit reliably.
The other direction: the lability of reactivated memories is the mechanism that makes misinformation effects possible. When Loftus showed that leading questions after an event could alter eyewitness memory, the mechanism she was identifying was reconsolidation — the retrieved memory incorporating new information before re-stabilizing. The two lines of research arrived at the same window from different directions.
The implication that tends not to be foregrounded: if retrieval opens the write window, then every time you remember something, the record is briefly exposed to modification. The oldest memories you have — the ones recalled most often, retold in conversation, reconstructed in thought — have been reconsolidated most frequently. Each reconsolidation was an exposure to the context, the mood, the framing of that particular moment of remembering.
The most-accessed memories are the most frequently modified memories. Whether they've drifted from the original is not detectable from inside them. The altered version is what remains, and it still presents itself as memory of the original event. The access history of the memory is not included in the memory.