Summary
- Experiment: Loftus and Palmer (1974) showed participants car crash videos and asked them to estimate speeds using different verbs like “hit” or “smashed.”
- Finding: When verbs like “smashed” were used instead of “hit,” people reported higher speeds and even remembered seeing broken glass that wasn’t there.
- Implication: The study revealed that memory is reconstructive, meaning it can be shaped and altered by the wording of questions.
- Application: This research highlighted the unreliability of eyewitness testimony and changed how police and courts handle questioning.
- Criticism: Critics argue the lab setting and use of video clips lack ecological validity, raising questions about how well the results apply to real-life events.
Psychologist Elizabeth Loftus has been particularly concerned with how subsequent information can affect an eyewitness’s account of an event.
Her main focus has been on the influence of (mis)leading information regarding both visual imagery and wording of questions concerning eyewitness testimony.
A leading question is a question that suggests what answer is desired or leads to the desired answer.
Loftus’ findings indicate that memory for an event that has been witnessed is highly flexible. If someone is exposed to new information during the interval between witnessing the event and recalling it, this new information may have marked effects on what they recall. The original memory can be modified, changed or supplemented.
This builds on Frederic Bartlett’s (1932) theory of reconstructive memory. Rather than replaying a fixed recording, people rebuild an event from memory, filling gaps using existing mental frameworks called schemas.
Bartlett’s own studies had participants repeatedly retell an unfamiliar Native American folk tale, the “War of the Ghosts”. Each retelling drifted further toward their own cultural expectations, a process he called effort after meaning. The same process operates in eyewitness memory.
Loftus and Palmer’s (1974) classic study, Reconstruction of Automobile Destruction, illustrates how leading questions can make eyewitness testimony unreliable.
Aim
Loftus and Palmer (1974) tested the hypothesis that the language used in eyewitness testimony can alter memory. They predicted that leading questions could distort eyewitness accounts, producing a confabulating effect in which cues in the question wording bias what a witness reports.
To test this, they asked people to estimate the speed of vehicles using different question wordings. Estimating vehicle speed is something people are generally poor at, so witnesses should be especially open to suggestion.
Experiment One: 5 verbs in leading questions
Procedure
Forty-five American students from the University of Washington formed an opportunity sample.
This was a laboratory experiment with five conditions, only one of which was experienced by each participant (an independent measures experimental design).
Seven short films of traffic accidents, taken from driver’s education and police training videos, were shown to participants.
The clips lasted between 5 and 30 seconds, with some staged crashes filmed at real speeds of 20, 30, or 40 mph.
Each group saw the films in a different order to avoid order effects. The whole session lasted about 90 minutes.
After each film, participants first gave a free account of what they saw, then answered the critical question.
“About how fast were the cars going when they (smashed / collided / bumped / hit / contacted) each other?”
Thus, the IV was the verb of the question, and the DV was the speed reported by the participants.
Findings
The estimated speed was strongly influenced by the verb used in the question.
The verb implied information about the speed, which systematically affected the participants’ memory of the accident.
| Verb | Mean Speed Estimate (mph) |
|---|---|
| Smashed (or “smashed into”) | 40.8 |
| Collided | 39.3 |
| Bumped | 38.1 |
| Hit | 34.0 |
| Contacted | 31.8 |
“Smashed” produced the highest average estimate (about 40.8 mph), followed by “collided” (39.3 mph), “bumped” (38.1 mph), “hit” (34 mph), and “contacted” (31.8 mph).
In reality, the filmed crashes were staged at 20–40 mph, showing that participants greatly overestimated the speeds. Statistical analysis confirmed the effect of wording was significant.
Conclusion
These results illustrate how the emotional intensity of the verb directly influenced participants’ speed estimates. More intense verbs like “smashed” produced significantly higher estimates than less intense verbs like “contacted”.
This suggests that eyewitness testimony can be biased simply by the way questions are worded.
Loftus and Palmer proposed two explanations:
- Response-Bias: The verb may simply have biased the number a person guessed, without creating a false memory of the crash.
- Memory Change: The verb may have altered what a person perceived and stored, so a more intense verb like “smashed” led them to remember a more serious accident.
To test this, Loftus and Palmer carried out a second experiment, designed to see whether leading questions simply bias responses or actually alter memory.
Experiment Two: The broken glass manipulation
In a second experiment, Loftus and Palmer tested whether leading questions simply bias responses or actually alter memory.
Procedure
150 students were shown a one-minute film which featured a car driving through the countryside followed by four seconds of a multiple traffic accident.
Afterward, the students were questioned about the film. The independent variable was the type of question asked.
- 50 participants were asked “how fast were the car going when they hit each other?”,
- 50 participants were asked, “How fast were the cars going when they smashed each other?”
- the remaining 50 participants were not asked a question about the car’s speed (i.e., the control group).
One week later, the dependent variable was measured – without seeing the film again, they answered ten questions, one of which was a critical one randomly placed in the list:
“Did you see any broken glass? Yes or no?”
There was no broken glass in the original film.
Findings
Participants who heard the cars had “smashed” into each other were more likely to report seeing broken glass a week later than those who heard “hit.”
The researchers found that the emotional intensity of the verb used in the initial questioning significantly influenced the probability that participants would falsely report seeing broken glass.
The results were as follows:
In the “smashed” condition, 32% of the participants (16 out of 50) incorrectly reported seeing broken glass.
In the “hit” condition, only 14% of the participants (7 out of 50) reported seeing broken glass.
In the control condition (who had not been asked about the speed), only 12% reported seeing broken glass.
Conclusion
The findings of Experiment 2 were critical for understanding how human memory operates. Loftus and Palmer had to differentiate between two potential explanations: a simple response bias, where a participant uses the verb as a guessing cue, or a genuine change in memory. The distinction matters.
The study demonstrated that a high-intensity verb could cause participants to vividly recall details that never actually occurred. Researchers therefore rejected the response-bias explanation in favor of genuine memory change.
The results support memory-as-reconstruction.
Memory is not a passive recording. It is an active process.
It is built from two sources, blending what was originally perceived with information supplied afterward, until the two become impossible to tell apart.
These findings are also highly consistent with schema theory, the idea that a mental framework of expectations shapes what we remember. The evocative verb “smashed” likely activated a schema for severe car accidents, one that carries an expectation of shattered glass.
As a result, the witnesses’ fragile memories were systematically distorted. They reconstructed the scene to include the broken glass that fit their activated schema. Even a single word in a question, it turns out, can permanently alter what a person remembers.
Confabulation
In cognitive psychology, confabulation is the process by which a person unintentionally introduces untrue elements into their recall of a past event. It often fills in missing information, or “gaps,” in memory.
Confabulation is not a deliberate lie or an attempt to deceive; rather, it is a natural characteristic of how human memory reconstructs the past.
The term is sometimes used for severe memory disorders, such as Korsakoff’s syndrome. Amnesic patients routinely guess to fill memory blanks. In Loftus and Palmer’s research, though, it refers to the everyday memory errors made by ordinary eyewitnesses.
Confabulation explains their second experiment’s results.
Participants who were asked how fast the cars were going when they “smashed” into each other were significantly more likely to falsely report seeing broken glass a week later. No glass was actually present in the film.
Psychologists differentiate between two related memory errors that occurred here:
- Distortion: Changing actual information to fit an existing mental framework (schema).
- Confabulation: Completely filling in missing information (gaps) through the use of a schema.
Human memory is not a passive video recording. It relies on active reconstruction. When a witness misses details, because of distraction or fear, the brain infers what must have happened from existing schemas, or mental expectations, of similar events.
When Loftus and Palmer used the emotionally intense leading verb “smashed,” it activated the participants’ schema for a severe car accident. The schema did the rest.
Because severe accidents typically involve shattered glass, the participants’ brains confabulated the broken glass to fill the gaps in their visual memory, making the memory feel complete and logical.
A Worked Example
Imagine two witnesses to the same minor car park bump. A week later, an officer asks one, “How fast was the car going when it smashed into the wall?” The other is asked, “How fast was it going when it contacted the wall?”
Neither witness has any new information. Only the officer’s verb differs. The first, whose “severe accident” schema is now active, may report a higher speed and later “recall” glass that was never there. The second, with no such schema primed, reports a lower speed and no glass.
One word did it. That is the everyday version of the effect Loftus and Palmer captured in the lab.
Loftus’s Wider Research Programme
The car crash study was the opening move in a decades-long programme showing how easily post-event information can distort memory (Loftus, 2005).
Loftus and Palmer’s car-crash paradigm tested how a question’s wording affects an estimate given moments later. Other studies in the programme show a similar pattern. Suggestion can add a false detail after a longer delay, or even implant a memory of an event that never took place at all.
Further Demonstrations of Memory Distortion
Several follow-up studies extended the same basic finding to other details and settings:
- Loftus and Zanni (1975): witnesses shown a car-accident film and asked whether they saw “the” broken headlight (a wording that presupposes it existed) said “yes” far more often than witnesses asked about “a” broken headlight.
- Loftus’s “barn” study (1975): Only 2.7% of witnesses asked about a Stop sign later claimed to see a nonexistent barn, versus 17.3% asked directly about “the barn.”
- Loftus, Miller and Burns (1978): Witnesses shown a Stop sign but later asked a misleading question about a Yield sign picked the correct slide only 41% of the time, against 75% for witnesses given accurate information.
- Loftus and Pickrell (1995), “lost in the mall”: with a family member’s help, participants were given a fabricated childhood story about being lost in a shopping mall; about 25% came to “remember” it, some with vivid, confident detail.
- Weapon Focus (Loftus, Loftus, & Messo, 1987): Witnesses who saw a gun remembered less about the culprit’s appearance than witnesses who saw the same person hand over a cheque, since attention narrows onto a threat during encoding.
Why Suggestion Distorts Memory
The leading explanation is source misattribution. Witnesses confuse information from a question with what they actually witnessed, so the two sources blend and become impossible to tell apart.
This general pattern, in which erroneous information supplied after an event reduces memory accuracy, is known as the misinformation effect.
Memon and Wright (1999) called this blending of separate sources memory blending, since witnesses cannot always tell which detail came from the original event and which came from the question.
Together, these studies show that memory distortion is not limited to speed estimates. Post-event suggestion can add a headlight, a barn, a road sign, or even an entire childhood event that never happened.
The common thread is that suggestion works best on peripheral, plausible details, not on facts a witness perceived clearly and knows to be true.
Strengths
The study’s main strengths are:
- High Control: A lab-based design held extraneous factors constant, isolating the verb as the only cause of change in estimates.
- Practical Implications: The findings shaped police and courtroom questioning guidelines and the cognitive interview.
- Replicable: The standardized procedure is easy to repeat, and the verb effect has been reproduced many times.
High level of control
Perhaps the greatest strength of Loftus and Palmer’s experiment is the degree of control over confounding variables.
Because the study was lab-based, the researchers could hold constant a range of factors, such as participants’ age, the incident viewed, and the environment. This let them ensure that only the verb condition was driving participants to re-evaluate their memories.
Practical Implications
The reconstructive memory hypothesis is extremely useful as a psychological explanation, for instance, in formulating guidelines for police questioning witnesses and suspects.
The conclusion that leading questions can affect memory has important implications for interviewing witnesses. It applies to police immediately after an event. It also applies to lawyers questioning witnesses much later in court.
Interviewers should avoid leading questions and be careful to word questions in a way that does not suggest an answer to the person they are interviewing.
The cognitive interview technique puts this into practice. It encourages free recall in the witness’s own words, from different perspectives and orders, while avoiding closed or leading questions.
The study also had real-world implications. Eyewitness misidentification is one of the leading causes of wrongful conviction. Based on evidence such as Loftus’s, the Devlin Report (1976) recommended that trial judges instruct juries that it is not safe to convict on a single eyewitness testimony alone.
Replicable
A strength of the study is it’s easy to replicate (i.e. copy). This is because the method was a laboratory experiment which followed a standardized procedure.
Weaknesses
The study’s main weaknesses are:
- Low Ecological Validity: Video clips lack the realism and consequences of a real accident, limiting generalization to real crimes.
- Biased Sample: Student participants are less experienced drivers than the general population, which may make them more easily swayed by wording.
- Demand Characteristics: cues that hint at a study’s purpose. Participants may have given the answers they thought the researchers wanted, rather than reporting a genuinely altered memory.
- Conflicting Research: Real-world studies such as Yuille and Cutshall (1986) found witnesses far more resistant to misleading questions than the lab predicts.
- Bounded Effect: Blatantly wrong or central, well-perceived details resist distortion (Loftus, 1979), so the effect is largely confined to peripheral, plausible information.
Low ecological validity
One limitation of the research is low ecological validity, the extent to which a study’s findings reflect what would happen in real-world settings. Participants viewed video clips, not a real-life accident.
A video clip lacks the emotional impact of a real crash, so participants may pay less attention and feel less motivated to be accurate. It also offers far less context: participants were cued to watch, whereas a real crash is unexpected.
There is no such warning in real life.
In an experiment, you expect to be asked about what you just watched, which may change how you attend to the film. In real life, your answers can carry consequences that add pressure no lab study does.
Overall, low ecological validity means the findings may not generalize well to how leading questions affect memory in real life.
Biased Sample
A further problem with the study was the use of students as participants. Students are not representative of the general population in several ways.
Differences between students and the broader population, such as age, memory abilities, learning habits, driving experience, and susceptibility to demand characteristics, could make it difficult to generalize the findings.
Importantly, they may be less experienced drivers and, therefore, less confident in their ability to estimate speeds. This may have influenced them to be more swayed by the verb in the question.
Demand Characteristics
Participants know they are in an experiment. This affects their behavior: they look for clues on how to behave, or demand characteristics, cues that hint at a study’s purpose. Most want to help the experimenter by giving the results they think are expected.
We cannot be sure memory was truly altered. Participants may simply have followed the researcher’s suggestions, in both the original questions and the follow-up.
Demand characteristics may have been “carried forward.” Having been asked about the cars “smashing,” participants may have been primed to report broken glass in the follow-up study.
Conflicting research
The strongest real-world challenge to Loftus and Palmer’s conclusions comes from a naturalistic study of an actual crime. Yuille and Cutshall (1986) found that misleading questions had far less effect on witnesses outside the laboratory.
Yuille and Cutshall (1986)
- Aim: To assess the accuracy of eyewitness accounts over time when subjected to misleading post-event information in a naturalistic setting.
- Procedure: Thirteen witnesses to a real armed robbery in Vancouver were interviewed four to five months later and compared against police reports, with two misleading questions (including a non-existent broken headlight) embedded in the interview.
- Findings: Misleading questions had virtually no distorting effect: all 13 witnesses retained accurate detail, and the most stressed witnesses were the most accurate.
- Conclusion: Reconstructive memory errors may be lab artifacts rather than true cognition failures: real trauma can trigger vivid, lasting flashbulb memories, leaving the original trace stable despite the misleading questions.
McCloskey and Zaragoza (1985)
In their second experiment, Loftus and Palmer concluded that misleading questions cause a genuine memory change, meaning the original memory is overwritten, altered, or destroyed. McCloskey and Zaragoza (1985) directly conflicted with this conclusion, proposing instead that Loftus and Palmer’s results were merely a product of response bias.
- Aim: To determine whether misleading post-event information permanently alters original memory traces or simply induces a response bias.
- Procedure: Participants saw slides of a theft including a hammer, then read a narrative wrongly calling it a wrench. A later test offered the hammer against a new screwdriver, omitting “wrench.”
- Findings: 66% of misled participants chose the correct hammer, against 69% of controls, a non-significant difference.
- Conclusion: Misleading narratives do not erase the original memory trace, since a destroyed memory would produce random guessing. The high retention shows the memory stayed intact, so the error reflects response bias, not genuine impairment.
Bekerian and Bowers (1983) found that when questions followed the events in strict sequential order, witnesses were not biased by later misleading questions. This suggests Loftus’s less structured questioning may have exaggerated suggestibility, though McCloskey and Zaragoza (1985) failed to replicate the specific ordering effect.
The Effect Has Limits
Not all misleading information is absorbed, however. When Loftus (1979) showed participants slides of a man stealing a red purse, 98% correctly identified its colour.
After reading a description that called it “brown,” all but 2% still remembered it as red. Well-perceived, central details resist distortion: the reconstructive effect operates mainly on peripheral, plausible details.
Contemporary Research
Loftus’s research programme remains active. Since 2015, the effect has extended to entirely fictitious events. Researchers have implanted rich false memories of committing a crime, reinforcing caution about “recovered” memories in therapy and police work.
The misinformation effect has also been studied in the context of fake news and social media. Repeated exposure to false claims shapes what people believe they witnessed.
Applied work has translated the original lesson into concrete reform. Sequential rather than simultaneous line-ups, double-blind administration, and recording a witness’s confidence immediately all reduce the risk that questioning contaminates memory.
The throughline is the same conclusion Loftus reached after thirty years of research: memory is malleable, and the wording of a single question can still help determine what someone recalls.
Independent Learning Tasks
- Draw a table showing the results of experiment one and draw a bar chart to show the results of experiment two.
- Read the original article of the study.
- Conduct your own study repeating one of the experiments by Loftus and Palmer.
- Use photographs (or video clips) of car accidents and write a set of questions, one of which will be the critical question.
- Test one group of participants using the “smashed” condition and the other group with the “hit” condition.
- Calculate the mean, median and mode speed estimates for both the “smashed” and “hit” conditions. Illustrate your results in either a table or graph.
Learning Check (1)
- Write an experimental hypothesis for experiment 1. Make sure it is clearly operationalized and include the independent and dependent variables.
- Why was it a good idea to ask 10 questions rather than just asking the critical question alone?
- Why was each group of participants shown the 7 video clips (of car accidents) in a different order?
- Outline the possible sampling technique that may have been used in this study.
- The participants knew they were taking part in a psychology experiment. How do you think this may have effected their behavior?
- Can you think of a way that this problem might of been overcome?
Learning Check (2)
- Write a null hypothesis for experiment 2. Make sure it is clearly operationalized and include the independent and dependent variables.
- What is a “control group”, and why is it necessary?
- What is an “experimental” group?
- Outline one difference between the responses given between the two experimental groups.
- Outline the quantitative measure used in this study.
- Why is it unsafe to convict a suspect based on a single eyewitness testimony?
References
Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge University Press.
Bekerian, D. A., & Bowers, J. M. (1983). Eyewitness testimony: Were we misled? Journal of Experimental Psychology: Learning, Memory, and Cognition, 9(1), 139–145. https://doi.org/10.1037/0278-7393.9.1.139
Devlin Committee Report: Report of the Committee on Evidence of Identification in Criminal Cases, 1976 Cmnd 338 134/135, 42
Loftus, E. F. (1975). Leading questions and the eyewitness report. Cognitive Psychology, 7(4), 560–572. https://doi.org/10.1016/0010-0285(75)90023-7
Loftus, E. F. (1979). Reactions to blatantly contradictory information. Memory & Cognition, 7(5), 368–374. https://doi.org/10.3758/BF03196941
Loftus, E. F. (2005). Planting misinformation in the human mind: A 30-year investigation of the malleability of memory. Learning & Memory, 12(4), 361–366. https://doi.org/10.1101/lm.94705
Loftus, E. F., Loftus, G. R., & Messo, J. (1987). Some facts about “weapon focus.” Law and Human Behavior, 11(1), 55–62.
Loftus, E. F., Miller, D. G., & Burns, H. J. (1978). Semantic integration of verbal information into a visual memory. Journal of Experimental Psychology: Human Learning and Memory, 4(1), 19–31.
Loftus, E. F., & Palmer, J. C. (1974). Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior, 13(5), 585–589. https://doi.org/10.1016/S0022-5371(74)80011-3
Loftus, E. F., & Pickrell, J. E. (1995). The formation of false memories. Psychiatric Annals, 25(12), 720–725.
Loftus, E. F., & Zanni, G. (1975). Eyewitness testimony: The influence of the wording of a question. Bulletin of the Psychonomic Society, 5(1), 86–88. https://doi.org/10.3758/BF03336715
McCloskey, M., & Zaragoza, M. (1985). Misleading postevent information and memory for events: Arguments and evidence against memory impairment hypotheses. Journal of Experimental Psychology: General, 114(1), 1–16. https://doi.org/10.1037/0096-3445.114.1.1
Memon, A., & Wright, D. B. (1999). Eyewitness testimony and the Oklahoma bombing. The Psychologist, 12(6), 292–295.
Yuille, J. C., & Cutshall, J. L. (1986). A case study of eyewitness memory of a crime. Journal of Applied Psychology, 71(2), 291–301. https://doi.org/10.1037/0021-9010.71.2.291
