Paper 12
The misinformation effect: editing the past after the fact
A whitepaper on how information encountered after an event can rewrite the memory of that event, the work on implanting whole false memories, why these errors happen, what they mean for eyewitness testimony, and how reconsolidation offers a candidate mechanism.
The one-sentence idea
A memory is not sealed the moment an event ends. What you learn about the event afterward · a leading question, a news report, a co-witness's account · can fold into the memory itself, so that later you recall the edited version and cannot tell it apart from what you actually saw. This is the misinformation effect, and its uncomfortable extension is that with enough suggestion an entire event that never happened can be installed as a confident personal memory.
The phenomenon is about as established as anything in experimental psychology. The mechanism is not settled, and this paper keeps those two things apart.
The founding experiment: the car crash
The cleanest way to see the effect is Elizabeth Loftus and John Palmer's 1974 study. Participants watched a film of a car accident, then answered questions about it. One group was asked how fast the cars were going when they "smashed" into each other. Another group got the same question with "hit" in place of "smashed." The single changed word raised the estimated speed. That alone shows wording steering a judgment.
The deeper result came a week later. Participants returned and were asked whether they had seen any broken glass. There was no broken glass in the film. The people who had earlier heard "smashed" were roughly twice as likely to report broken glass that was never there. The post-event word had not merely biased a speed estimate in the moment. It had reached back into the stored memory of the scene and added a detail consistent with a worse crash.
That is the signature of the misinformation effect: information supplied after the event becomes part of the remembered event, and the person cannot flag it as an insertion.
The standard laboratory version, refined across many studies (Loftus, Miller and Burns, 1978, is the canonical one), has three phases. A person witnesses an event, usually a slideshow or film. Some time later they receive a narrative or a set of questions that contains a misleading detail, for example a stop sign described as a yield sign. Later still they are tested on the original event. Reliably, a fraction of witnesses now report the misleading detail as though they had seen it. Evidence tag: established, and robust across decades and laboratories.
Escalation: implanting whole memories
If a false detail can be added, can a whole false event be built? Loftus and Jacqueline Pickrell's "lost in the mall" study (1995) tested this. With help from a family member, researchers gave each participant short accounts of childhood events, three true and one fabricated: getting lost in a shopping mall around age five, frightened and eventually rescued by an elderly stranger. Over repeated interviews, about a quarter of participants came to "remember" the mall episode, some adding vivid details of their own that were never supplied.
Later work pushed further. Studies have led sizable minorities of participants to recall spilling punch on the bride's parents at a wedding, being attacked by an animal, or, in one striking line of research (Shaw and Porter, 2015), coming to recall committing a crime that brought them into contact with police. The exact rates are debated and depend heavily on method. Shaw and Porter reported that a large majority of their participants met their criterion for a false crime memory, but independent researchers who recoded the same reports arrived at substantially lower figures, so the headline number should be read with caution. The qualitative finding, however, is not seriously disputed: suggestion, repetition, and imagination can manufacture confident autobiographical memories of events that did not occur. Evidence tag: established as a phenomenon, contested in its magnitude and in how far the strongest crime-memory claims generalize.
The distinction matters. Adding a false detail to a real memory (the broken glass) and constructing an entirely false episode (the mall) may not be the same operation. Both are real; treating them as one thing overstates what any single study shows.
Why it happens: source monitoring
The leading explanation is not that memories are simply overwritten and erased. It is that memory keeps poor records of where its contents came from. This is the source-monitoring framework (Johnson, Hashtroudi and Lindsay, 1993).
Every remembered detail has an origin: something you saw, something you were told, something you imagined or inferred. Memory stores the content far more reliably than the tag that says where it came from. When you later retrieve "broken glass," the content is there, but the label distinguishing "saw it in the film" from "inferred it from the word smashed" is weak or gone. A plausible, vivid, internally generated detail gets misattributed to perception. The error is a source-monitoring error: right content, wrong origin.
This reframes false memory as ordinary memory operating on incomplete metadata rather than as a special malfunction. It also explains the conditions that make misinformation stick. Suggestions that are plausible, that are repeated, that the person is asked to imagine vividly, and that come from a trusted source are exactly the ones whose origin is hardest to keep straight. The following table summarizes what moves the effect.
| Factor | Direction | Why, in source-monitoring terms |
|---|---|---|
| Plausibility of the suggested detail | More plausible, more accepted | Fits the gist, so a generated detail looks perceptual |
| Repetition of the misinformation | More repetition, stronger effect | Familiarity is misread as evidence of having seen it |
| Delay before the misleading input | Longer delay, larger effect | Original trace weakens, leaving less to contradict the suggestion |
| Trust in the source | More authority, more uptake | An authoritative account is scrutinized less for its origin |
| Warning that misinformation may come | Warning reduces the effect | Prompts deliberate source checking at retrieval |
The companion whitepaper on arousal and false memory develops a related route, gist-driven reconstruction, where meaning-consistent details are supplied because they fit the scene. Source monitoring and gist reconstruction are complementary accounts of the same family of errors, not rivals.
The stakes: eyewitness reliability
None of this would carry much weight if it stayed in the lab. It does not. The eyewitness who saw a real crime is then interviewed by police, talks to other witnesses, sees media coverage, and views a lineup, all of which are post-event information delivered exactly when the original memory is weakening and most suggestible.
The applied lesson is that confidence is not a reliable index of accuracy. A witness can be certain, detailed, and emotionally invested in a memory that misinformation has edited, precisely because the edit is invisible from the inside. The person is not lying. The source tag that would let them separate what they saw from what they were later told is simply not available to introspection.
This work reshaped procedure. It is a major reason for non-leading interview protocols, double-blind lineup administration where the officer running the lineup does not know the suspect, and recording of a witness's confidence at the first identification before contamination accumulates. The scientific claim, that memory is reconstructive and suggestible, is established; the reforms are its practical translation.
Connecting to reconsolidation
Now the bridge to the rest of this corpus, offered as a candidate mechanism and labeled as such. The connection below is an analogy between two literatures, not a demonstrated identity.
The reconsolidation story (companion whitepaper one) says that recalling a settled memory can return it briefly to a labile state, during which what re-stabilizes may differ from what was recalled. The misinformation paradigm, viewed through that lens, looks like reconsolidation in the wild. The witness reactivates the memory of the event, either by being questioned or by mentally replaying it. During the interview, new information arrives. If retrieval has reopened the trace, that new information is not merely stored alongside the original; it can be written into the reconsolidating trace, which is exactly why the edit later feels like an original perception rather than a later addition.
Two points from the reconsolidation literature make this more than loose analogy. First, the window is thought to open on prediction error, on retrieval carrying a mismatch. A leading question that introduces a discrepancy is a natural way to generate that mismatch. Second, the phenomenon in animals is genuinely mechanistic (Nader, Schafe and LeDoux, 2000), which is what makes the human bridge tempting.
But the caution has to be equally plain. The misinformation effect is fully explainable at the cognitive level by source monitoring alone, with no reconsolidation required. The reconsolidation account is a candidate mechanism for why some post-event edits become permanent and indistinguishable, not a demonstrated cause of the classic effect. The animal-to-human translation gap applies here as everywhere in this corpus: the molecular story is clean in rodents and inferred, not measured, in a person misremembering a stop sign. Treat the connection as a hypothesis that unifies two literatures, not as an established finding.
What to take away
- Post-event information rewrites memory for an event. Loftus and Palmer's "smashed versus hit" study showed a single word later producing false reports of broken glass that was never in the film.
- The effect scales up. The lost-in-the-mall work and its successors show that repeated suggestion can implant whole false autobiographical events in a substantial minority of people. The phenomenon is established; its magnitude and its most dramatic crime-memory extensions are contested.
- The leading mechanism is source monitoring. Memory stores content more reliably than the tag for where the content came from, so a suggested or imagined detail gets misattributed to perception.
- Confidence is not accuracy. An edited memory feels exactly like a true one from the inside, which is why eyewitness procedure was reformed toward non-leading interviews and blind lineups.
- Reconsolidation is a candidate mechanism, not a proven one. The misinformation effect is fully explained cognitively without it, and the animal-to-human gap keeps the neural bridge speculative.
Sources
- Loftus and Palmer (1974). Reconstruction of automobile destruction: an example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior.
- Loftus, Miller and Burns (1978). Semantic integration of verbal information into a visual memory. Journal of Experimental Psychology: Human Learning and Memory.
- Loftus and Pickrell (1995). The formation of false memories. Psychiatric Annals.
- Shaw and Porter (2015). Constructing rich false memories of committing crime. Psychological Science.
- Johnson, Hashtroudi and Lindsay (1993). Source monitoring. Psychological Bulletin.
- Nader, Schafe and LeDoux (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature.
Evidence tags (established, emerging, contested) follow the corpus convention and keep the animal-to-human translation gap visible.