Why You Remember It Differently Now

Ulric Neisser, whose 1967 book gave cognitive psychology its name, carried a memory of the day Pearl Harbor was attacked. He was listening to a baseball game on the radio. The broadcast was interrupted with the news, and he ran upstairs to tell his mother. He held that memory for decades before the obvious thing occurred to him. The attack was on 7 December. Nobody broadcasts baseball in December. He published the memory, and the catch, in 1982, in a collection he edited, in a chapter called “Snapshots or benchmarks?”

Four years later he got the chance to test it on other people. On the morning of 29 January 1986, the day after the Challenger exploded, Neisser and Nicole Harsch at Emory University handed a questionnaire to a large psychology class and asked where the students had been when they heard, who told them, what they were doing. About two and a half years on they asked the same students the same questions. About 40 per cent of the accounts had changed, and about a quarter of the students had changed every detail they were asked about. Their confidence had not moved. It bore no relation to whether they were right.

That result has been repeated on a larger scale since. A consortium of American researchers surveyed people within a week of the September 2001 attacks and again at 11 months, 25 months and ten years. Memories of where people were and how they heard decayed fast in the first year and then stopped decaying. Confidence stayed high the whole decade. And the inconsistent versions, once they appeared, tended to be repeated at the next survey rather than corrected.

Those studies describe what happens. The work on why came twenty years after the Challenger study, from animal research.

What reconsolidation is

In 2000 Karim Nader, Glenn Schafe and Joseph LeDoux at New York University trained rats to fear a tone by pairing it with a shock, and then waited for the memory to settle. A new memory is known to be fragile for a few hours; block protein synthesis in the amygdala during that window and the memory does not form. After that it was assumed to be fixed. Nader’s group played the tone again, a day or two weeks later, so that the rat recalled the shock, and injected the protein synthesis blocker straight after. The rats lost the fear. The same drug given without the reminder left the memory intact, and the same drug given six hours after the reminder did nothing. Calling the memory up had made it soft again for a few hours, and in those hours it had to be rebuilt from new protein or it was gone.

They called it reconsolidation.

In 2007 Almut Hupbach and colleagues at the University of Arizona moved the idea into something closer to a human life. Students learned a list of objects on day one. On day two, some were given a subtle reminder of that session before learning a second list; others just learned the second list. On day three, asked to recall the first list, the students who had been reminded mixed in objects from the second list. The unreminded students did not. The intrusion did not happen if the test came straight away on day two; it needed the overnight gap, which is what you would expect if the reminded memory had been reopened and was being rebuilt with whatever was to hand. And only the reminded list was affected. The second list, never reopened, stayed clean.

That is the mechanism the Challenger students had been demonstrating. Every time one of them told the story of where they were, they opened the file. Whatever was in the room when they opened it, the version a friend told, the footage they had watched since, a detail that made the story better, had a chance of being written back in. Recalling a memory reopens it, and it does not close in the state it was opened in.

What it was supposed to fix

If a fear memory goes soft when it is recalled, then a phobia or a trauma could be recalled on purpose and rewritten while it is open, with no drug involved. In 2010 Daniela Schiller, with LeDoux and Elizabeth Phelps among the co-authors, published a human version of this in Nature. People were conditioned to fear a coloured square paired with a shock. The next day the square was shown once, as a reminder, and then shown repeatedly without shocks, the standard way of extinguishing a fear, but inside the window the reminder had opened. The fear did not come back the following day, and a year later it still had not, while in people who had the extinction without the reminder, it did.

That paper is widely cited and it is where the pop version comes from: every time you remember, you rewrite, and therapy can use that.

In 2020 Anastasia Chalkia, Tom Beckers and colleagues at KU Leuven ran a preregistered, high-powered, direct replication of the critical conditions in Schiller’s first experiment. They found no benefit of the reminder over ordinary extinction. The fear came back either way. Along the way they also found that the exclusion criteria in the original paper had been misreported, which held their data collection up for a long time while they worked out what had actually been done. Their conclusion was that it is open to question whether there is robust evidence that the procedure prevents the return of fear in people at all.

Two years earlier James Elsey, Vanessa van Ast and Merel Kindt in Amsterdam, a group that has itself run many of the propranolol studies, had gone through the whole human literature and set out a framework for what would count as showing reconsolidation and what would not. Their verdict on the human evidence was that there is a wealth of observations consistent with it, a number that are not, and enough alternative explanations that reconsolidation cannot be conclusively inferred at the level of the brain from what has been shown so far. They called it a viable and hotly contested explanation.

The evidence, then, is uneven. In rats, under tight conditions, a recalled memory can be chemically disrupted. A reminder can let new material leak into a student’s old list. And the trick that was supposed to rewrite a conditioned fear in people did not survive a direct retest. The pop claim, that memories are rewritten each time you recall them and this can be harnessed, runs well ahead of the second and third of those.

What is settled, and what the Challenger students still show

What holds is older than the word reconsolidation and does not depend on it. Memory is reconstructive. Neisser knew it from his own December baseball game, and Frederic Bartlett had shown it in Cambridge in the 1930s with people retelling a folk tale and smoothing it toward what made sense to them. The flashbulb studies add the part that matters for a life: the reconstruction does not feel like one. Confidence detaches from accuracy early and stays detached. The wrong version, once told, is the version that gets told again.

Reconsolidation adds a mechanism for how the telling does the damage, and the mechanism is well supported in animals and reasonably supported in the Arizona lists. Whether it can be run in reverse, on purpose, to take the fear out of a memory, is not established, and the best evidence from a direct replication says the flagship demonstration does not hold.

What none of it gives you is a way to tell, from inside, which of your memories have been through this. Neisser caught his because a calendar contradicted it. The Challenger students were caught by a questionnaire they had filled in the morning after. Most memories carry nothing like that. The ones you are surest of are the ones you have told most often, which on Hupbach’s result means they are the ones that have been reopened most often, with the most occasions for something else to be written in. The research can show that this happens. It cannot show you where, in your own case, it has.