The Memory That Rebuilds Itself

A. For most of the twentieth century, memory was imagined as a form of storage. Experience was thought to leave a trace — a physical mark somewhere in the brain — and remembering was the act of locating that mark and reading it. The metaphors of the age reinforced the idea: memory was a filing cabinet, then a tape recording, then a hard drive. Each comparison assumed the same thing, that the stored item sits unchanged until retrieved.

B. The evidence against this view accumulated slowly and then all at once. In the 1970s, the psychologist Elizabeth Loftus showed participants a film of a car accident and afterwards asked them how fast the cars were going when they smashed into each other. Others were asked the same question with the word hit. Those given the more violent verb reported higher speeds, and — more strikingly — were later more likely to report having seen broken glass in the film. There was no broken glass. A single word in a question, asked after the event, had altered what people believed they had witnessed.

C. Findings of this kind were initially treated as curiosities about eyewitness unreliability, of interest mainly to lawyers. Their deeper implication took longer to be accepted: that the act of remembering is not passive. Each time a memory is brought to mind it appears to become temporarily unstable, and is then written back in a slightly altered form. The technical term is reconsolidation. Under this model, a memory recalled a hundred times is not a well-preserved memory but a heavily edited one.

D. The clinical possibilities were noticed quickly. If a memory must be rewritten each time it is recalled, then interfering during that window might change it permanently. Trials with people suffering from severe trauma have tested exactly this, asking patients to recall a distressing event and then administering a drug that dampens the emotional response before the memory settles again. Results have been mixed but not negligible; in several studies the memory remained accessible while its emotional charge was measurably reduced. The event was still known, but it no longer arrived with the same force.

E. Such work raises questions that laboratories are not equipped to settle. A person shaped by a painful memory may not straightforwardly want it softened, and the boundary between relieving suffering and editing a life is not one that any experiment can locate. Practitioners in the field have generally been careful to frame the aim as reducing intrusive distress rather than removing the past — a distinction that matters more in principle than it is easy to maintain in practice.

F. Meanwhile the everyday consequences are largely unappreciated. Confidence in a memory turns out to correlate only weakly with its accuracy: people are frequently certain about details they have unknowingly constructed. Nor does repetition help. Telling a story often makes it smoother and more coherent, and coherence is easily mistaken for truth. The memories we trust most — the ones we have told and retold — are precisely the ones that have passed through the editing process most often.

G. None of this makes memory worthless. A system that reconstructs rather than replays is, for most purposes, more useful than a perfect archive would be; it generalises, discards the irrelevant and updates itself as circumstances change. The error lies not in the mechanism but in our expectations of it. We were never given a recording device. We were given something closer to a storyteller, and storytellers revise.