You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Eyewitness testimony (EWT) is the account given by a person who witnessed a crime or significant event, and it has long carried enormous weight in criminal investigation and in court. A confident witness pointing across a courtroom and saying "that is the man" is one of the most persuasive things a jury can hear. Yet decades of psychological research show that this confidence is often misplaced: human memory is not a video recording that faithfully replays the past, but a reconstructive process that actively rebuilds an event each time it is recalled — and that reconstruction can be distorted by the way a witness is questioned, by information encountered after the event, by the presence of a weapon, and by the witness's own emotional state. Because mistaken eyewitness identification has contributed to real, documented wrongful convictions, understanding why and how EWT goes wrong is not an academic exercise but a matter of justice. This lesson examines the reliability of EWT through four factors — reconstructive memory, leading questions and post-event information, weapon focus, and anxiety — and studies in depth the classic experiment that founded the modern field, Loftus and Palmer (1974) on leading questions and speed estimates.
Key Definition: Eyewitness testimony (EWT) is the account a witness gives of a crime or event, used as evidence. It depends on the accuracy of memory, which — because memory is reconstructive — can be altered by the wording of questions, by information received after the event, and by the witness's emotional state at the time.
This lesson addresses the reliability of eyewitness testimony content of Edexcel 9PS0 — Paper 2, Topic 6: Criminological Psychology, specifically the factors that affect the accuracy of EWT (reconstructive memory, leading questions and post-event information, weapon focus and anxiety) and the classic study of Loftus and Palmer (1974) on the effect of leading questions on the recall of a car accident.
Connects to…
The single most important idea underpinning the whole topic is that memory is reconstructive, not reproductive. The everyday intuition — that remembering is like pressing "play" on a faithful internal recording — is wrong. When we witness an event, we encode a partial, fragmentary trace; when we later recall it, we actively rebuild the event from that trace, filling gaps using schemas (organised packets of prior knowledge and expectation about how the world works) and using whatever information happens to be available at the moment of retrieval, including the way a question is phrased. This reconstructive view descends from Bartlett's early work on how people distort unfamiliar material to fit their existing schemas, and it explains why EWT is systematically vulnerable: if remembering involves rebuilding, then the materials used in the rebuild — expectations, later information, the interviewer's wording — can all leak into the final "memory" without the witness being aware of it.
Key Definition: Reconstructive memory is the process by which recall involves actively rebuilding a memory from fragments, using schemas and currently available information, rather than replaying a stored, unaltered record. This makes memory efficient but also prone to distortion.
This is why a witness can be entirely honest and entirely confident yet wrong. They are not lying; they are faithfully reporting a memory that has been reconstructed with contaminating material woven into it. This dissociation between confidence and accuracy is one of the most consequential findings in the field, because juries — reasonably but mistakenly — treat a witness's confidence as a guide to their reliability.
Misleading information is incorrect or suggestive information supplied to a witness after the event, which may alter their memory of it. It takes two principal forms: leading questions and post-event information (or discussion).
Key Definition: A leading question is a question phrased so as to suggest a particular answer, or which contains an embedded assumption that may influence the response. "How fast was the car going when it smashed into the other?" implies a violent, high-speed collision before the witness has said anything.
The power of leading questions lies in the fact that the question itself becomes part of the information the witness uses to reconstruct the event. A single emotive verb, or a grammatical presupposition, supplies a detail that the reconstructive process may absorb.
Key Definition: Post-event information is information encountered after witnessing an event — from other witnesses, from questioning, or from the media — that becomes incorporated into the witness's own memory of what happened.
When witnesses to the same crime discuss it, they frequently absorb details from one another that they did not personally see — a phenomenon called memory conformity. Because each witness observes the event from a different position, discussion allows one witness's account to contaminate another's. Research on co-witness discussion (for example, Gabbert and colleagues, 2003) has shown that a large proportion of witnesses who discuss an event will subsequently report details supplied only by a co-witness, whereas witnesses who do not confer make this error far less often. The practical implication is direct and important: police should obtain statements from witnesses separately and promptly, before they have a chance to confer.
Memory conformity can arise through two routes, which map exactly onto the social-influence topic: informational social influence (the witness assumes the other person's account must be correct, especially when uncertain) and normative social influence (the witness wants to agree and fit in with the group of witnesses).
A further factor is the presence of a weapon. The weapon-focus effect describes the tendency for a witness's attention to be drawn towards a weapon at the expense of other details — most damagingly, the perpetrator's face, which is exactly the detail investigators most need.
Key Definition: The weapon-focus effect is the reduction in accuracy of eyewitness identification that occurs when a weapon is present, because the witness's attention narrows onto the weapon and away from other features such as the culprit's appearance.
The classic demonstration is Johnson and Scott (1976). Participants waiting for a study overheard a conversation in an adjacent room. In the low-anxiety condition a man then emerged holding a pen, with grease on his hands; in the high-anxiety condition a man emerged holding a blood-stained knife. When participants later tried to identify the man from a set of photographs, identification was less accurate in the knife condition (around a third correct) than in the pen condition (around half correct). The interpretation is that the anxiety created by the weapon narrowed attention onto the knife, impairing memory for the culprit's face.
There is, however, an important alternative explanation. Pickel (1998) argued that a weapon is not only threatening but also unusual and unexpected in most settings, and showed that attention narrows onto items that are surprising or out of context (in one study, a raw chicken) just as much as onto threatening ones. This raises the possibility that weapon focus reflects surprise/unusualness, a confounding variable, rather than anxiety specifically — a point examined further in the evaluation.
The effect of anxiety on EWT is genuinely contradictory. Some studies show anxiety impairs memory (as in the weapon-focus finding); others show it enhances memory. The standard way of reconciling the two strands is the Yerkes–Dodson law.
The Yerkes–Dodson law proposes an inverted-U relationship between physiological arousal (anxiety) and performance: too little arousal and the witness is under-alert and inattentive; a moderate level of arousal produces optimal attention and recall; but very high arousal overwhelms the witness and recall declines again.
flowchart LR
LOW["Low anxiety<br/>under-aroused<br/>(poor recall)"] --> MOD["Moderate anxiety<br/>optimal arousal<br/>(BEST recall)"]
MOD --> HIGH["High anxiety<br/>over-aroused<br/>(poor recall)"]
On this view, whether anxiety helps or harms depends on where on the curve a given witness sits — which is precisely why laboratory studies of sudden, extreme fear and real-life studies of moderate, sustained stress can reach opposite conclusions.
Set against the laboratory impairment findings, studies of real crimes often find high anxiety associated with better memory. Witnesses to an actual shooting (studied by Yuille and Cutshall, 1986) were interviewed months after the event and found to be remarkably accurate, and resistant to misleading questions the researchers deliberately inserted; those who reported the highest stress at the time were, if anything, the most accurate. Similar findings emerged from studies of witnesses to real bank robberies, where those most directly threatened (the victims) tended to show the best recall. These real-world findings directly contradict the weapon-focus impairment result and cast doubt on how far laboratory studies generalise.
A refinement that helps reconcile the conflict is the distinction between central and peripheral detail. Emotionally significant, central features of an event (the assailant's actions, the weapon itself) tend to be well retained and may even be sharpened by high arousal, whereas peripheral details (background, bystanders, surrounding objects) are more readily lost or distorted — a pattern sometimes called tunnel memory. This helps explain why real-crime witnesses can be highly accurate about the core of a terrifying event while weapon-focus studies find impaired memory for the peripheral detail of the culprit's face. It implies the blunt question "does anxiety help or harm EWT?" is too crude: the answer depends on which kind of detail is recalled.
The study that founded the modern experimental analysis of EWT is Loftus and Palmer (1974), "Reconstruction of Automobile Destruction." It comprised two linked experiments.
Aim: to test whether the verb used in a question about a car accident affects witnesses' estimates of the speed the cars were travelling.
Procedure: forty-five American students watched several film clips of traffic accidents. After each clip they completed a questionnaire that included one critical question: "About how fast were the cars going when they ___ each other?" Participants were divided into five groups, each receiving a different critical verb: smashed, collided, bumped, hit or contacted.
Findings: the verb systematically affected the mean speed estimate. The most violent verb, "smashed", produced the highest estimate (around 40.8 mph), and the mildest, "contacted", produced the lowest (around 31.8 mph) — a difference of roughly 9 mph, despite every group watching the same clips.
| Critical verb | Mean speed estimate (mph) |
|---|---|
| Smashed | 40.8 |
| Collided | 39.3 |
| Bumped | 38.1 |
| Hit | 34.0 |
| Contacted | 31.8 |
Conclusion: a single word in a question can bias an eyewitness's report. But this could reflect either response bias (the verb influences the answer without changing the underlying memory) or genuine memory distortion. Experiment 2 was designed to distinguish these.
Aim: to test whether a leading question distorts the actual stored memory or merely biases the immediate response.
Procedure: a new sample of 150 students watched a film of a car accident. One group was asked the speed question using "smashed", a second using "hit", and a third (control) was not asked about speed at all. One week later, with no further viewing of the film, all participants returned and were asked: "Did you see any broken glass?" There was no broken glass in the film.
Findings: participants who had earlier heard "smashed" were more than twice as likely as the "hit" group to falsely remember seeing broken glass a week later.
| Condition | Reported seeing broken glass |
|---|---|
| "Smashed" | Highest (~32%) |
| "Hit" | Lower (~14%) |
| Control (no verb) | Lowest (~12%) |
Conclusion: because the false detail (broken glass) appeared at a later test, one week after the misleading verb, the leading question appears to have altered the stored memory itself — Loftus's substitution or memory-replacement account — rather than merely biasing the immediate response. The wording did not just change what participants said; it changed what they remembered.
graph TD
A["Witness views event"] --> B["Original memory trace<br/>(fragmentary)"]
B --> C{"Leading question<br/>e.g. 'smashed'"}
C --> D["Post-event information<br/>absorbed into trace"]
D --> E["Reconstructed memory<br/>(now includes false detail:<br/>'broken glass')"]
style E fill:#e74c3c,color:#fff
style B fill:#27ae60,color:#fff
Loftus and Palmer converted a plausible worry — that questioning could distort testimony — into a rigorously demonstrated, replicable experimental effect, with a clear causal manipulation (a single word) and a clean design. It provided the empirical foundation for reforms in police interviewing, and it is the touchstone study for the reconstructive view of memory: if a single verb can implant broken glass into a witness's memory a week later, then the "memory as video recording" model cannot be right.
The specification rewards understanding not just that misleading information distorts EWT but how. There are competing accounts, and distinguishing them is a route to higher marks.
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.