Published on 2026-07-25
Can a single word in a question change what you remember?
In a 1974 study, one verb in a question about a car crash made witnesses recall broken glass that was never in the film. The word rewrote the memory.
Two witnesses watch the exact same short film of a car accident. Afterward, one is asked how fast the cars were going when they contacted each other. The other is asked how fast they were going when they smashed into each other. Same footage, same four seconds of screech and impact. Different word. And a week later, the second witness will be far more likely to swear they saw broken glass scattered on the road. There wasn't any. Not in that clip, not in any version of it.
That is the finding of a study run in 1974 by Elizabeth Loftus and John Palmer at the University of Washington. In the first part of the experiment, 45 students watched short films of traffic collisions taken from driving safety courses, then answered questions about what they'd seen. The critical question always took the same shape, but the verb changed: 'about how fast were the cars going when they smashed / collided / bumped / hit / contacted each other?' Nothing else about the question was different. Just that one word.
One verb, ten extra miles an hour
The estimates moved with the verb, cleanly and consistently. People asked about cars that 'smashed' guessed an average speed around 40 mph. People asked about cars that merely 'contacted' each other guessed closer to 32 mph. Same crash. An eight mile an hour gap created entirely by word choice. The films themselves showed collisions staged well below even the lowest of those guesses, so every group was already overestimating; the verb just decided how far past reality each group drifted.
The glass that was never filmed
A speed estimate being off by a few miles an hour might feel like a rounding error, the sort of thing that happens whenever someone guesses at numbers. So Loftus and Palmer ran a second version of the experiment with 150 students to test something sharper: not an estimate, but a supposed fact. A week after watching the crash, participants were asked a plain yes or no question: did you see any broken glass? There was none in the film. Among those who had originally been asked about cars that 'smashed', 32 percent said yes, they remembered glass. Among those asked about cars that merely 'hit', only 14 percent said yes. A group who'd never been asked a speed question at all landed at 12 percent, close to the baseline rate of people simply guessing wrong.
That gap is the whole point. The word 'smashed' didn't just nudge a number upward. It planted an idea of violence severe enough to shatter glass, and a week later that idea had been absorbed into what people believed they had witnessed with their own eyes. Loftus went on to give this general phenomenon a name: the misinformation effect, a memory error that appears when a person is exposed to incorrect or suggestive information somewhere between an event and whatever moment they're later asked to recall it, whether that moment is a police interview, a courtroom, or a conversation with a friend.
The question asked about a memory doesn't just probe it. Sometimes it becomes part of it.
Beyond the lab
This matters well past psychology courses, because the mechanism that worked on 45 college students in a lab in 1974 works the same way in a police interview, a family retelling of an old dispute, or a group chat rehashing what 'really' happened at a party. Hundreds of follow-up studies since Loftus and Palmer's work have confirmed the broader pattern: memory can be contaminated by information introduced after the fact, and the person absorbing that information rarely notices the seam. They don't feel like they're being told what to remember. They feel like they're remembering.
None of this means every eyewitness account is worthless, or that questions are inherently dangerous. It means the phrasing of a question is never neutral, even when nobody intends any harm by it. An interviewer who says 'did you notice the man's weapon' has planted a weapon into the conversation regardless of whether one existed. A friend who retells your own story back to you with one intensified detail can leave you a little more certain of a detail you never actually saw.
The fix isn't to distrust every memory you have. It's to notice, especially in anything that matters, how the story was first put into words: yours, or somebody else's, and whether the version you're now sure of was shaped as much by a question as by whatever actually happened in front of you.
Sources
- Loftus and Palmer (1974): Car Crash Experiment (Simply Psychology) (simplypsychology.org)
- Eyewitness Testimony and Memory Biases (Noba Project) (nobaproject.com)