Misinformation effect is not quite what it seems.

A single word in a question can plant a memory of something that never happened.

In 1974, Elizabeth Loftus and John Palmer showed participants films of car accidents and asked them to estimate the cars’ speed. One group was asked how fast the cars were going when they hit each other. Another group got the same question with a single word changed: smashed. The smashed group estimated significantly higher speeds. A week later, both groups were asked whether they remembered seeing broken glass in the films. There was no broken glass. The smashed group was more than twice as likely to say they remembered it.

This is the misinformation effect: exposure to misleading information after an event distorts the memory of the event itself. The new information does not sit alongside the original memory as a separate, questionable claim. It infiltrates and rewrites. Participants were not lying about the broken glass. They genuinely remembered it, because the word smashed had supplied wreckage their minds were happy to render.

Loftus spent the following decades mapping the effect’s reach, and it is vast. Leading questions reshape memories of stop signs into yield signs. Post-event suggestions alter recall of faces, actions, and sequences. In the most dramatic demonstrations, entirely false childhood events, being lost in a mall, taking a hot-air balloon ride, could be planted in a substantial minority of participants through suggestive interviewing, complete with emotional detail and confident belief. The machinery that updates memory with new information cannot distinguish correction from corruption.

The mechanism is reconstructive memory. Recalling is not replaying a recording; it is rebuilding the past from fragments plus current knowledge, and the rebuilding process happily incorporates whatever materials are at hand, including things learned after the fact. Each retelling, each question, each news report is another ingredient in the reconstruction. Memory is not a transcript. It is a draft, and it is always accepting edits.

The stakes made this the most applied finding in memory science. Eyewitness testimony, once the gold standard of evidence, was shown to be exquisitely vulnerable to the wording of police questions, to line-up procedures, to post-event media. Reforms followed: blind line-up administration, careful question wording, video recording of interviews. The Innocence Project’s exonerations, hundreds of them, with mistaken eyewitness identification a factor in the large majority, are the misinformation effect written in human cost.

Imagination alone can do the damage. In 1996, Garry and colleagues demonstrated imagination inflation: simply imagining a childhood event that never happened, like breaking a window, increased participants’ confidence that it had really occurred. The imagined scene, rehearsed in detail, acquired the characteristics the source-monitoring system reads as real: vividness, sensory detail, emotional texture. The brain’s reality test was passed by a fabrication, because the test checks qualities, not history.

The most unsettling demonstration came from Loftus and Pickrell’s lost-in-the-mall study. Through suggestive interviews, about a quarter of participants developed detailed false memories of being lost in a shopping mall as children, complete with elaborations their families confirmed never happened. They were not confused or compliant. They remembered. The study drew ethical fire, and rightly so, but its scientific point stands: the machinery that builds true memories will build false ones from the same materials, and it will not label the difference.

Loftus published the smashed-versus-hit experiment in 1974. Half a century later, in an age of viral claims and suggestive feeds, the finding reads less like a laboratory curiosity and more like a description of the information environment. Every question you are asked about the past is a small edit to your memory of it. Choose your questions, and your sources, accordingly.