The broken glass that wasn’t there: how memory rebuilds

Elizabeth Loftus and the implanting of memories. · published 9 September 2026

Under the right conditions a person can be given a detailed, sincerely held memory of something that never happened — which says something about how memory is built, and much less than it is often used to say about whether to believe someone.

The angle

In Loftus and Palmer’s 1974 study there was no broken glass in the film at all — the people who “remembered” it were the ones who had been asked how fast the cars were going when they “smashed”, which shows memory is assembled at the moment of recall from fragments and whatever the question implies, not replayed from a recording.

The notes this episode was written from

These are the working notes, unedited. Everything the video states as fact comes from here, and nothing else may be stated as fact. The sections on limits and on what the finding does not show are part of the brief, not an afterthought — see the method.

The early work: Loftus and Palmer (1974). Participants watched a film of a car accident and were asked how fast the cars were going when they “smashed”, “collided”, “bumped”, “hit” or “contacted” each other. The verb changed the speed estimates. A week later, those who had heard “smashed” were more likely to report having seen broken glass in the film. There was no broken glass.

The famous one: Loftus and Pickrell (1995), usually called “Lost in the Mall”. Twenty-four participants were given short accounts of four childhood events, supplied by a family member. Three were real. The fourth — being lost in a shopping centre as a young child, and helped by an elderly stranger — was invented. About a quarter came to report remembering it, some adding details of their own that nobody had supplied.

Later studies implanted more elaborate events. Wade and colleagues (2002) used doctored photographs to give people memories of a childhood hot-air balloon ride that never happened. Others have produced memories of spilling punch at a wedding, or of being attacked by an animal.

Scoboria and colleagues (2017) pooled eight of these studies and found that roughly a third of participants showed signs of a false memory, and around a quarter accepted the event as having happened.

What it establishes

Memory is reconstructive rather than a recording. It is assembled at the moment of recall from fragments, expectation and whatever the question implies — and confidence in a memory does not track its accuracy. This work reshaped how courts treat eyewitness testimony.

Limits

The 1995 study had twenty-four participants and its exact rate should be given as “about a quarter” rather than a precise figure. Brewin and Andrews (2017) argued the rate of full, richly recalled false memories is lower than the headline numbers suggest, and that many counted cases are partial — the person accepts the event happened without actually remembering it. The conditions are also specific: a trusted family member vouching for the event, and repeated suggestion over several sessions.

What this does not show

It does not show that memory is generally unreliable, that most memories are invented, or that people who report abuse are not to be believed. The finding is that memories can be created under particular, engineered conditions — not that they usually are. This research has been used in court to discredit testimony, Loftus’s expert witness work is genuinely contested, and the topic sits over real allegations by real people. The script must not imply that someone’s account of their own life is suspect, and must not use “false memory” as a euphemism for lying: the whole point is that the people in these studies were sincere.