Memory and the Eyewitness: Why Honest People Report Strange Things

Memory is not a recording that can be played back. Each time we remember something, the brain rebuilds it from fragments, filling the gaps with what we expect, what we know and what we have learned since. Usually the result is good enough. Under the wrong conditions it can be confidently, sincerely wrong.

One of the first people to show this was the Cambridge psychologist Frederic Bartlett. In his 1932 book Remembering he described asking people at Cambridge to read an unfamiliar Native American folk tale called "The War of the Ghosts" and then retell it, either repeatedly over time or passed from one person to the next like a game of telephone. With each retelling the story grew shorter and more conventional. Details that did not fit the readers' expectations were dropped or turned into something more familiar, while the general gist survived. Bartlett concluded that remembering is an act of reconstruction shaped by what we already know.

This matters most for strange experiences. The stranger an event is, the less a memory can lean on familiar patterns, and the more it gets filled in with expectation, including expectations drawn from films, folklore and the reputation of a place. None of this is lying. It is how everyone's memory works, believers and skeptics alike, and understanding it is the first step to weighing a strange report fairly.

The psychologist Elizabeth Loftus has spent decades showing how easily memory can be changed after the fact. In a classic 1974 study with John Palmer, students watched films of car crashes and were asked how fast the cars were going when they "hit" or "smashed into" each other. Those who heard "smashed" gave higher speed estimates. A week later, asked whether they had seen any broken glass, which the films did not show, 32 percent of the "smashed" group said yes, compared with 14 percent of those asked with "hit".

Loftus and others went on to show that misleading information given after an event can be absorbed into the memory of the event itself, a finding known as the misinformation effect. In the 1995 "lost in the mall" study, Loftus and Jacqueline Pickrell gave volunteers short accounts of childhood events supplied by relatives, plus one false account of being lost in a shopping mall. About a quarter came to remember the false event, at least in part. A preregistered replication published in 2023 found a similar result, with about a third doing so.

The practical lesson is about questions. Asking "Did you see the figure in the doorway?" plants a figure and a doorway; asking "What did you see?" does not. Confidence is also a poor guide once a memory has been retold and questioned many times, because retelling makes a memory feel more certain without making it more accurate. Good investigators ask open questions first and write down the answers before offering any ideas of their own.

Expectation affects what people remember seeing, not just how they describe it. In a 2003 study in the British Journal of Psychology, Richard Wiseman, Emma Greening and Matthew Smith held fake séances in which an actor playing a medium suggested that a table had levitated, although it had not moved at all. Afterward about a third of the participants reported that the table had moved. Believers in the paranormal were more likely than disbelievers to say so, and in a second experiment believers were more susceptible to suggestion only when the suggestion fitted their belief. About a fifth of participants thought the fake séances had included genuine paranormal phenomena.

The study echoes a long history. Nineteenth-century séances took place in darkness, with sitters primed to expect contact and a medium narrating what was supposedly happening, ideal conditions for suggestion to shape what people afterward sincerely remembered.

The lesson for a skeptic is not that believers are gullible. Expectations work on everyone, and any of us remembers more readily what fits what we already think, which is why careful researchers try to check their own assumptions as hard as anyone else's. The lesson for a believer is that a sincere memory of a moving table is not, on its own, evidence that the table moved. A camera running throughout, or an independent observer who was told nothing in advance, is worth more than any number of confident recollections.

Strange encounters tend to happen in the worst possible conditions for accurate perception: at night, in poor light, by surprise and often in fear. In dim light the eye relies on rod cells, which do not register color and give poor detail. The center of the retina, where vision is sharpest in daylight, has no rods at all, which is why astronomers look slightly to one side of a faint star to see it. Peripheral vision is good at detecting movement and poor at identifying shapes, so a glimpse "out of the corner of the eye" is exactly where the brain does the most filling in.

Attention matters as much as light. In a famous 1999 experiment, Daniel Simons and Christopher Chabris asked viewers to count basketball passes in a short video. About half of them failed to notice a person in a gorilla suit who walked through the middle of the scene. The effect, called inattentional blindness, shows that we see what we attend to and can miss even striking things we are not looking for. The reverse is true as well: when we are primed for something, such as a figure in a reputedly haunted hallway, we find it more readily.

Fear narrows attention further and makes time feel stretched, so frightened witnesses often remember events as lasting longer than they did. None of this means a witness saw nothing. It means the conditions of an encounter are part of the evidence and should always be recorded alongside the account.

Several witnesses sound stronger than one, and sometimes they are. But witnesses who talk before giving their accounts are less independent than they appear. In a 2003 study in Applied Cognitive Psychology, Fiona Gabbert, Amina Memon and Kevin Allan showed pairs of people different videos of the same event, each containing details the other had not seen, and let some pairs discuss what they had watched. Afterward, 71 percent of the witnesses who had discussed the event mistakenly reported at least one detail they could only have learned from their partner. Psychologists call this memory conformity.

The same blending happens outside the laboratory. Members of a family compare notes after a frightening night. Visitors to a famous location have read the same stories and watched the same videos. A witness gives a local newspaper an interview, and the published version becomes the one everyone, including the witness, remembers. Over time an account can also grow, as stories do in retelling: it becomes sharper, simpler and more dramatic, and acquires a satisfying ending, the same process that shapes urban legends.

For anyone collecting reports, the practical rule is simple. The most valuable account is the earliest one, written down by each witness separately before they have compared notes or read anything about the place. Later accounts still matter, but they should be read with an eye to what may have flowed in from conversation and coverage.

A principle often quoted in these debates is that extraordinary claims require extraordinary evidence. Carl Sagan made the phrase famous on television in 1980, but the sociologist Marcello Truzzi had already written that extraordinary claims require extraordinary proof, and the idea is older still: in 1814 Pierre-Simon Laplace argued that the more extraordinary a fact, the stronger the evidence it needs, and in 1748 David Hume wrote that a wise man proportions his belief to the evidence. Critics point out that what counts as extraordinary is partly a matter of opinion, and Truzzi himself warned against what he called pseudo-skeptics, critics who deny a claim without accepting any burden of proof for their denial.

In practice, both sides can work through the same questions:

• When was the account first written or recorded, and by whom? Was that before or after the witness talked to others or read about the place?
• What were the conditions: light, distance, duration, and the witness's state, such as tired, frightened or just woken?
• Are there witnesses who gave their accounts independently, before comparing notes?
• Is there evidence beyond memory, such as an original photo file, a recording or a measurement?
• What ordinary explanations fit, and what would each one predict that can be checked?
• What evidence would change the believer's mind, and what would change the skeptic's?

Honest people report strange things all the time. Understanding memory is not a way to dismiss them. It is a way of taking their accounts seriously enough to test them.