Can we trust our memories?

Why do memories change each time we remember them?

A 2014 U.S. National Academies report, Identifying the Culprit, sums up decades of research: unknown to the person, memories are “forgotten, reconstructed, updated, and distorted.” Classic evidence comes from Frederic Bartlett’s 1932 book Remembering. He had people read a Native American story, “The War of the Ghosts,” and retell it later. Because the story came from a culture unlike their own, they changed details they could not understand. Some later partial repeats failed. But in a 1999 study, Erik Bergman and Henry Roediger followed his method closely. People forgot parts of the story and also reshaped it, more so as weeks and months passed, and some added new material. Bartlett’s finding held up.

Why might recalling change a memory? One idea is reconsolidation: a recalled memory may turn fragile again and need storing anew. In a 2000 study, rats that recalled a fear memory lost it when a drug that blocks protein-making was put into their brains right after; given six hours later, or without the recall, the drug left it intact. In people, the evidence is mixed: a 2020 registered replication could not reproduce a much-debated human result built on this idea. We label reconsolidation Supported, still debated: clear in a key animal study, mixed in people.

How can someone clearly remember something that never happened?

What we hear later can blend in. In Elizabeth Loftus and John Palmer’s 1974 study, 45 students watched films of car crashes. Asked how fast the cars were going when they “smashed” into each other, they guessed about 40 miles per hour on average; asked with “contacted,” about 32. In a second experiment with 150 students, 32% of the “smashed” group said a week later that they had seen broken glass, against 14% of the “hit” group. A 2014 meta-analysis notes hundreds of studies of this misinformation effect, and found that warning people afterward cut it to less than half its size, on average. But a 2017 test with a hockey collision found that calling the hit a “contact,” “bump” or “smash” made no difference to speed estimates, suggesting the word effect may depend on the setting.

Whole events can be suggested. In the influential “lost in the mall” study of 1995, about a quarter of people came to report a false memory of being lost in a shopping mall as a child. In 2023, a team in Ireland repeated the study with 123 people, five times the original sample, and a plan registered in advance. The team coded 35% as reporting a false memory, and mock jurors were highly likely to believe those events had happened.

How common are such memories? Researchers disagree. A 2017 mega-analysis of eight studies (423 people) classed 30.4% of cases as false memories. But a 2017 review by Chris Brewin and Bernice Andrews found some memory-like experience in 47% of people on average but likely full memories in only about 15%, and concluded that we are less open to such false memories than has been suggested.

Are vivid, confident memories more accurate?

Psychologists call vivid memories of hearing shocking news flashbulb memories. On September 12, 2001, 54 Duke University students wrote down how they had first heard about the attacks, and also a recent ordinary event. Retested 1, 6 or 32 weeks later, both memories had become less consistent at the same pace, but only the ordinary one lost its vividness and the students’ belief in its accuracy. The authors concluded that flashbulb memories are special in how accurate they feel, not in how accurate they are.

A larger study surveyed more than 3,000 people in seven U.S. cities within a week of the attack and sent follow-up surveys over ten years. Their memories of how they heard the news faded fast in the first year, then leveled off, while confidence stayed high. Some intense memories do stay steady, though. In a 2007 study, 72 people answered the same questionnaire about their near-death experiences almost two decades apart, with no significant change; the researcher concluded the accounts had not been embellished. See What are near-death experiences, and what do they tell us?

Can hypnosis uncover hidden memories?

Studies do not support that. In 1985, a report in JAMA, the Journal of the American Medical Association, concluded that memories drawn out under hypnosis can include invented details and false memories. They were not more accurate than ordinary recall, and appeared less reliable. A 1994 meta-analysis of 24 studies found that hypnosis helped accuracy after 1 to 2 days but hurt it after less than a day or a week, and brought more errors, more false memories and more confidence. Because confidence rises along with errors, a memory that feels clearer after hypnosis is no sign that it is true. A 2025 review found some evidence that hypnosis can aid memory, with contradictory results, and urged caution with hypnotic regression and guided imagery, which may lead to false memories. We label the claim that hypnosis recovers accurate hidden memories Not supported.

Why can honest eyewitnesses still get things wrong?

The National Academies’ report lists conditions that shape what a witness sees: dim light, a brief look, a long distance, stress, strong emotion, and a weapon such as a gun or knife. Police procedures can shape memory too. The report tells of Jennifer Thompson, a college student who was sexually assaulted in North Carolina in 1984. She struggled to choose between two photos, then picked Ronald Cotton, and detectives told her, “You did great.” In a later live lineup, Cotton was the only person repeated from the photos, and by the trial she was “absolutely sure.” DNA tests cleared Cotton after 10.5 years in prison; the two later wrote Picking Cotton together. As of April 2020, the Innocence Project counted eyewitness misidentification in 252 of 367 U.S. DNA exonerations (69%). It stresses that mistaken eyewitnesses are not lying; they sincerely believe they picked the right person.

There are fixes. The report recommends double-blind lineups, run by someone who does not know which person is the suspect. It also urges standard instructions, video recording, and writing down the witness’s confidence, in their own words, at the first identification. In a 2017 review, John Wixted and Gary Wells found that when the first test is fair and uncontaminated, a witness’s high-confidence pick is remarkably accurate, but unfair lineups can undermine even a confident one.

So can we trust our memories, and how can you check one?

The strongest objection is that everyday memory usually works well. In a 2020 study, adults recalled real events that researchers could check, days to years later. They recalled less as time passed, but 93% to 95% of the details they did recall were accurate, far above what memory scientists and other academics guessed in a survey. In a 2020 review, Chris Brewin, Bernice Andrews and Laura Mickes argued that three claims behind the popular view that memory is unreliable rest on “shaky foundations,” and that there is no reason to drop the old consensus that memory is malleable but essentially reliable. Both fit the evidence: memory is usually right about what we recall freely, and suggestion, leading questions and feedback can change it.

This matters for anyone whose memories shaped their life: a loved one’s last words, a presence felt after a death, a light in the sky you still cannot explain. Memory research does not say such memories are false, but details can shift, and, as Loftus and Palmer noted, most people are markedly inaccurate at reporting time, speed and distance, which matters for sightings. See Is it normal to feel that someone who died is still with me? and Have UFOs been shown to be alien spacecraft? If a memory of loss weighs on you, our care page may help.

What do people often get wrong?

“Memory works like a video camera.” It rebuilds rather than replays, though Brewin and colleagues question how widely people really hold the camera view. “If I’m sure, I’m right.” Confidence can grow with feedback and repeat lineups, as Jennifer Thompson’s case shows. “Memory can’t be trusted at all.” For what we recall on our own, it usually can.

Check it yourself

• Write it down early, before talking it over with others.
• Separate what you saw from what you concluded. Note the plain facts first, then your interpretation.
• Watch the questions. Ask “What happened?” rather than “How fast did they smash?”
• Check against records. Dated photos, messages and news reports can confirm or correct a memory. For clips and images, see How can I check a viral video before I share it?
• Notice when confidence grows. Feeling surer after being told you got it right, or after being asked again, is not new evidence; see our guide to common thinking traps.

The answer stays the same: memory is mostly reliable for what we recall freely, it can shift over time and under suggestion, and honest people can be wrong.