How does surprise rewrite old memories?

Here, surprise means a mismatch between a predicted outcome and the outcome that actually occurs. Researchers call that mismatch prediction error. It does not require a dramatic emotional reaction. A familiar sound followed by an unexpected absence can create a prediction error in a learning experiment.

The central question is what the mismatch means. If a warning signal stops predicting danger, perhaps the old association needs revision. But perhaps this particular situation is an exception, and the warning remains useful elsewhere. Updating an existing association changes an established expectation; learning an exception preserves it while limiting when it applies.

This helps explain why a contradiction is not automatically evidence that the past was wrong. New experience might reveal a changed world, a different context, or an earlier misunderstanding. The learning problem includes deciding which explanation fits.

In fear-conditioning experiments, researchers pair a tone with an unpleasant event. An animal may then show a defensive response to the tone itself. Researchers can repeatedly present the tone without the unpleasant outcome, a procedure called extinction.

If the response declines, it is tempting to conclude the association has been deleted. Yet fear can return after time passes, after exposure to the unpleasant event, or when the context changes. Bouton’s review of extinction research describes the importance of context: an animal may learn that a signal is safe under particular conditions while the earlier danger association remains influential elsewhere.

The careful conclusion is that extinction frequently includes new, context-sensitive learning rather than straightforward erasure. A response disappearing today is not proof that the underlying association can never matter again.

In a 2024 eLife study, Nicholas Kennedy and colleagues trained rats to associate a tone with a mild shock, then compared ways of removing it. During standard extinction, the tone occurred without the shock. During gradual extinction, shock intensity was progressively reduced first—a series of smaller mismatches instead of one abrupt one.

Both reduced freezing during extinction. But under the tested conditions, rats given gradual extinction showed less later recovery of conditioned fear. The researchers interpreted this as support for a model in which smaller prediction errors favor updating an existing memory state, while larger discrepancies may encourage a separate state alongside the old one.

These were measurements of conditioned animal behavior, not rats revising personal stories. The results do not establish that gradual change always updates memory or that a large surprise always creates a separate memory.

In a 2000 experiment, Karim Nader, Glenn Schafe, and Joseph LeDoux reactivated rats’ tone-related fear memories and then blocked protein synthesis in the amygdala. Later expression of the fear memory was impaired; the same intervention without reactivation did not produce the same loss. This helped establish the idea of reconsolidation.

But remembering is not automatically a rewrite operation. Not every retrieval destabilizes a memory. A critical review by Elsey and colleagues found many observations consistent with reconsolidation while emphasizing inconsistent results and alternative interpretations. Evidence for memory modification is not evidence for a reliable method of removing unwanted human memories.

A 2021 study by Alyssa Sinclair and colleagues had participants learn short narrative videos. Later, some familiar videos stopped before their expected endings. The interruptions affected later recall, including incorrect details on delayed tests, and brain-imaging analyses linked some effects to changes in hippocampal activity patterns.

A 2026 correction reported a preprocessing error; after reanalysis, several numbers and figures changed, and the researchers reported that the principal findings remained unchanged.

A surprising event can draw attention while also changing how earlier information is reconstructed. A revised memory may feel compelling, but that feeling does not establish accuracy. Surprise is a reason to examine the evidence, not a guarantee that the newest version is true.

The studies concern conditioned fear, short videos, and controlled experiments, so they do not hand out rules for daily life. They do suggest a way of thinking. When something contradicts what you expected, your memory system may be more open to updating than it is during routine moments.

That openness cuts both ways. A surprising correction can help replace an outdated belief, but a surprising misleading suggestion could also slip in. Checking the source of a startling new detail before folding it into an old memory is a sensible habit.

The extinction findings also offer a reassuring note. A fear or habit that seems to return is not necessarily a sign that earlier progress was fake. Old learning and new learning can sit side by side, and the context can tip which one shows up.