You are viewing a free preview of this lesson.
Subscribe to unlock all 9 lessons in this course and every other course on LearningBro.
Elizabeth Loftus and John Palmer's 1974 study is the classic study for the cognitive-area theme of memory, and it is one of the most consequential experiments in the whole of psychology — the study that, more than any other, established that human memory is not a faithful video recording but an active reconstruction that can be reshaped by the very questions we are asked about it. Its finding is easy to state and unsettling in its implications: simply changing one word in a question about a filmed car crash — asking how fast the cars were going when they "smashed" rather than "hit" each other — changed people's estimates of the speed, and, a week later, changed whether they "remembered" seeing broken glass that was never actually there.
This lesson tells the study in the OCR "tell the story" format: the background that motivated it, the aim, the method (design, sample and step-by-step procedure of both experiments), the results with their real figures, Loftus and Palmer's conclusions, and a full evaluation of its method, data, ethics, validity and reliability. It closes by linking the study to its key theme, its area, the relevant perspectives and the debates it fuels. Because Loftus & Palmer anchors so many synoptic answers — on the reliability of eyewitness testimony, on reconstructive memory, on the usefulness of cognitive research — knowing it precisely is one of the highest-yield investments you can make for Component 02.
| This lesson covers | OCR H567 Component 02 element | AO focus |
|---|---|---|
| Background: reconstructive memory and the reliability of eyewitness testimony | Section A — Cognitive; theme: memory (classic) | AO1 knowledge |
| Method: two laboratory experiments; samples (n and who); the verb manipulation and glass follow-up | Section A — Core study (Loftus & Palmer) | AO1; AO2 |
| Results (mean speed estimates per verb; the broken-glass figures) | Section A — Core study | AO1 |
| Conclusions (memory as reconstructive; response-bias vs memory-substitution) | Section A — Core study | AO1; AO3 |
| Evaluation: method, data type, ethics, validity, reliability, sampling, ethnocentrism | Section A; Section B debates | AO3 |
| Links to theme, area (Cognitive), perspective and debates | Section B — Areas, perspectives, debates | AO1; AO3 |
The specification is referenced descriptively; consult the official OCR H567 specification document for its exact published wording. This lesson develops AO1 (the aim, procedure, results and conclusions), AO2 (applying reconstructive-memory ideas to novel scenarios of eyewitness error) and AO3 (evaluating the study's validity, reliability, ethics and generalisability).
By the early 1970s, psychologists influenced by Bartlett's earlier work on remembering had grown increasingly convinced that memory does not work like a tape recorder. Rather than storing an exact copy of an event and replaying it on demand, the mind was thought to reconstruct the past each time it is recalled — assembling a plausible account from fragments of the original experience, from general knowledge and expectation (schemas), and from information encountered after the event. If that view is correct, then memory should be malleable: the way we are questioned about an event should be able to alter what we "remember" of it. Bartlett's own "War of the Ghosts" studies had shown that when people recall an unfamiliar story they unconsciously alter it to fit their existing expectations — dropping unfamiliar details, rationalising the strange, and importing culturally familiar elements. Loftus & Palmer can be read as taking Bartlett's insight out of the study of remembered stories and into the far more consequential arena of remembered events, with a manipulation precise enough to satisfy the newly rigorous, experimental cognitive psychology of the 1970s.
Loftus and Palmer set out to test this idea in a domain of enormous practical importance: eyewitness testimony. Courts routinely treat a confident eyewitness as powerful evidence, on the common-sense assumption that memory faithfully records what happened. But if a witness's memory can be distorted merely by the wording of the questions a lawyer or police officer asks, then eyewitness testimony is far more fragile than the legal system assumes. The researchers focused specifically on leading questions — questions whose phrasing suggests or contains information that can bias the answer — and asked whether such questions could reshape a memory of a traffic accident. The distinction they were probing is the difference between a question that merely retrieves a memory and one that edits it: everyone accepts that a badly phrased question might elicit a poor answer, but the radical claim is that the question could reach back and change the underlying record itself.
The choice of a car accident as the target event was deliberate and shrewd. It is an event ordinary people feel confident they can judge (everyone has an opinion about vehicle speeds), yet it is precisely the kind of fast, unexpected, emotionally charged incident about which real eyewitnesses are questioned — and about which, the researchers suspected, memory is in fact poor and therefore open to influence. Speed estimation in particular is notoriously unreliable: most people, including experienced drivers, are poor at judging the velocity of vehicles, so the "true" answer is fuzzy and there is plenty of room for a cue in the question to nudge the estimate. Understanding why the researchers expected memory to be malleable clarifies the significance of the result: they were not testing whether people can be tricked into a wrong answer, but whether the underlying memory representation itself can be altered by post-event language. This is why the study is usually cited as the foundational demonstration of the misinformation effect — the general finding that information encountered after an event can contaminate the memory of the event itself.
The overarching aim was to investigate whether the language used in a question about a witnessed event — specifically a leading question — can distort a person's memory of that event. The two experiments pursued this aim in stages:
Both were laboratory experiments using an independent-measures (independent-groups) design: different participants were allocated to the different verb conditions, so that no participant heard more than one version of the critical question. The independent variable in each experiment was the critical verb used in the question about the accident; the dependent variables were the estimated speed (Experiment 1) and the report of broken glass (Experiment 2).
Sample. Participants were 45 students, divided into five groups of nine. They were an opportunity sample of American college students.
Procedure. Participants watched seven short film clips of traffic accidents, taken from driver-safety teaching films. After each clip they completed a questionnaire that asked them to give an account of the accident and then answered a series of questions, among which was one critical question about the speed of the vehicles. This critical question was worded identically for all participants except for one verb. The question was, in essence: "About how fast were the cars going when they [verb] each other?" — where the verb was one of five, differing in the severity of collision they imply:
| Group | Critical verb | Implied severity |
|---|---|---|
| 1 | smashed | most severe |
| 2 | collided | severe |
| 3 | bumped | moderate |
| 4 | hit | mild (the "neutral" baseline) |
| 5 | contacted | least severe |
Each group of nine received only one verb. The dependent variable was the mean speed estimate (in miles per hour) given by each group. Because the same film clips were shown to every group, any difference in speed estimates could be attributed to the verb alone — a clean experimental logic. The design controls potential confounds carefully: the films were identical across conditions, the questionnaire was identical except for the one word, and the critical question was embedded among fillers so its purpose was not obvious. The independent-measures structure was essential — had one participant seen both "smashed" and "hit" versions, they would likely have noticed the manipulation and adjusted their answers, so allocating each participant to a single verb protected against that awareness.
The purpose of Experiment 2 was to determine whether the verb had genuinely altered the memory of the event or had merely biased the response at the moment of questioning. This is the theoretically decisive experiment, because Experiment 1 on its own is compatible with a sceptical reading: perhaps the verb simply nudged an uncertain guess without touching the memory at all. To adjudicate between "the verb changed the answer" and "the verb changed the memory", Loftus and Palmer needed a later test of a specific, checkable detail that the verb had never mentioned — hence the broken-glass measure, taken a full week after the event, about a detail that was demonstrably absent from the film.
Sample. A new, larger sample of 150 students, again an opportunity sample of American college students, divided into three groups of fifty. Using a fresh, larger sample here (rather than re-testing the Experiment 1 participants) avoided any carry-over from the first study and gave more stable frequency counts for the false-memory measure.
Procedure. Participants watched a single film clip of a multiple-car accident lasting about four seconds (the collision portion was brief). They then answered a set of questions about the accident, including the critical speed question. The critical manipulation was again the verb in the speed question:
The control group is the design's quiet masterstroke. Without it, a difference between "smashed" and "hit" would tell us only which verb produced more glass reports, not what the baseline rate of false glass reporting is. By including a group never asked a leading question, Loftus and Palmer could show that the "hit" group behaved essentially like the control while the "smashed" group departed upward from both — locating the effect specifically in the severity-implying verb rather than in the mere act of being asked about speed.
The crucial measure came one week later, when all participants returned and answered a further set of ten questions without seeing the film again. Embedded among these — again, not flagged as special — was the key question: "Did you see any broken glass?" — to which the correct answer was no, because there was no broken glass in the film. The dependent variable was the number of participants in each group who reported (wrongly) that they had seen broken glass. The week-long delay was deliberate: it made an immediate response-bias explanation implausible (no fresh "smashed" cue was present at the glass question) and allowed any altered memory to consolidate, so that what was measured was the stored recollection rather than a snap judgement.
The verb systematically affected the speed estimates: the more severe the collision implied by the verb, the higher the estimate. The mean speed estimates were approximately as follows:
| Critical verb | Mean speed estimate (mph) |
|---|---|
| smashed | 40.8 |
| collided | 39.3 |
| bumped | 38.1 |
| hit | 34.0 |
| contacted | 31.8 |
The gap between the extreme verbs is the headline: participants asked about cars that "smashed" estimated roughly 41 mph, while those asked about cars that "contacted" estimated roughly 32 mph — a difference of about 9 mph produced purely by changing one word in the question, even though every group had watched the same clips.
The verb manipulation carried over into a false memory of a detail a week later:
| Group (verb in week 1) | Reported seeing broken glass | Did not report broken glass |
|---|---|---|
| "smashed" (n = 50) | 16 | 34 |
| "hit" (n = 50) | 7 | 43 |
| Control — no speed question (n = 50) | 6 | 44 |
Participants who had been asked the "smashed" question a week earlier were more than twice as likely to report seeing broken glass (16 of 50) as those asked the "hit" question (7 of 50) or the control group (6 of 50) — despite there being no broken glass in the film. The verb from a week earlier had planted a detail that was never there.
Loftus and Palmer drew two connected conclusions.
First, and most directly, the wording of a question can influence the answer a witness gives: a leading question containing a verb that implies greater severity produced higher speed estimates. This alone shows that eyewitness testimony is sensitive to how questions are framed.
Second, and more profoundly, the broken-glass result led them to argue that the leading question does not merely bias the answer at the moment of questioning but can alter the stored memory representation itself. They distinguished two possible explanations of the effect:
The broken-glass finding — that a week later the "smashed" group falsely remembered glass — is hard to explain by momentary response-bias alone, because there was no immediate answer being biased; it points instead toward the memory-substitution account. Loftus and Palmer therefore concluded that memory is reconstructive: the original perception of the event and the information supplied after the event (here, the loaded verb) are combined into a single, altered memory. This is a landmark demonstration of the misinformation effect — the distortion of memory by post-event information — and a direct challenge to the reliability of eyewitness testimony.
The reconstructive-memory account they invoke is worth spelling out, because it is the study's real theoretical payload. On this view, remembering is not retrieval of a stored copy but an act of building: at recall the mind fuses whatever fragments of the original experience remain with inferences drawn from schemas (general knowledge of how car crashes go) and with any post-event information absorbed. The verb "smashed" activates a schema of a violent, high-speed collision, whose default features — high speed and, crucially, broken glass — can be written into the memory as though perceived. The participant who "remembers" glass is not lying or consciously guessing; they experience the reconstructed memory as genuine. This is why the effect is so troubling for the law: a sincerely confident witness and an accurate one can be indistinguishable from the inside.
It is important to read the conclusion precisely, because it is easy to over-state. The study does not show that all eyewitness memory is worthless, nor that every witness can be made to remember anything. It shows something narrower and better evidenced: that memory for a fast, briefly-seen event is malleable, and that the language of subsequent questioning can shift both an immediate judgement (speed) and a later recollection of a specific detail (glass). The drama of the study lies in how little it took — a single verb — to produce the distortion, and in the fact that participants were not lying or guessing wildly but sincerely reconstructing what they believed they had seen.
The real-world eyewitness-testimony implications follow directly. If a single verb in an otherwise neutral question can plant a false detail, then investigative questioning becomes a potential source of contamination: a question such as "did you see the knife?" (rather than "a knife") risks not merely eliciting a mistaken answer but rewriting the witness's memory before it is heard in court. Repeated interviews, post-crime media coverage and conversation between co-witnesses all supply post-event information that the reconstructive account predicts will blend into the remembered event. The study therefore reframes eyewitness memory as a fragile record to be protected from contamination at every stage between event and courtroom.
The laboratory design gave Loftus and Palmer exceptional control and standardisation: every group saw the same film clips and answered the same questionnaire, differing only in the critical verb, so a change in the dependent variable can be attributed to the verb with confidence. This gives the study high internal validity (for the effect of the verb on the response) and makes it highly replicable — and it has indeed been replicated and extended many times. The trade-off is artificiality, discussed under validity below.
The study yielded quantitative data — mean speed estimates in mph and frequency counts of broken-glass reports — which are objective and readily compared across conditions. This is a strength for reliability and analysis. The limitation is that a numerical speed estimate is a fairly narrow operationalisation of the rich concept of "memory distortion", and the study captures little about why individual participants answered as they did.
By the standards of the more notorious social studies, Loftus & Palmer is relatively benign, but it is not ethically neutral:
Subscribe to continue reading
Get full access to this lesson and all 9 lessons in this course.