Case File 078 · Investigation and Psychology

How Stress Affects Testimony

High stress tends to make eyewitness memory less accurate, and visible stress in an interview is not a reliable sign of deception.

By Murder Book Editorial TeamPublished 16 September 2026Updated 16 September 2026

High stress generally makes testimony less accurate, not more. A 2004 meta-analysis by Deffenbacher, Bornstein, Penrod and McGorty combined dozens of experiments and found that high stress reduced both the accuracy of eyewitness identifications and the accuracy of witnesses' recall of crime details. A related line of research on "weapon focus" shows that when a weapon is present, witnesses tend to remember less about other parts of the scene, including the person holding it. Stress affects the other side of the interview too. A nervous witness or suspect is not necessarily lying, because anxiety is a normal response to being questioned about something serious. The practical lesson is that vivid, frightening experiences are not automatically remembered accurately, and visible stress is not a reliable sign of deception.

Key takeaways

A 2004 meta-analysis found that high stress impaired both eyewitness identification and recall of details.
Weapon focus research shows a weapon's presence tends to reduce memory for other elements of the scene.
A study of military survival school trainees found poorer recognition of an interrogator under high stress than under lower stress.
A frightening event can feel vivid and certain in memory and still be recalled inaccurately.
Signs of stress during an interview are not proof of lying; honest people are often anxious too.

What does the research say about stress and eyewitness memory?

The most cited summary is the meta-analysis by Kenneth Deffenbacher and colleagues, published in Law and Human Behavior in 2004. They analysed 27 independent tests of stress on identifying a perpetrator or target person, and 36 tests of stress on recalling details of a crime.

The results strongly supported the idea that high stress harms both kinds of memory. The overall effect sizes were the same, -.31, for correct identifications and for the accuracy of recall. The effects were not uniform:

Condition in the studies	Effect of high stress
Target-present line-ups	Notably larger than for target-absent line-ups
Eyewitness identification studies	Larger than for face recognition studies
Studies using a staged crime	Larger than for studies inducing stress in other ways

The pattern in the last row matters. Studies that most closely resemble a real crime showed the largest effects, which suggests laboratory research may, if anything, understate the problem.

Is there evidence from situations more stressful than a laboratory?

Researchers cannot ethically frighten people the way a real crime can, so a study from military training is often cited. Charles Morgan and colleagues at Yale studied 509 active-duty military personnel attending survival school, where training includes realistic mock interrogations.

According to Yale's announcement of the study, the high-stress condition followed 48 hours of food and sleep deprivation and involved physical confrontation during interrogation. In the lower-stress condition, the interrogator tried to trick participants into revealing information. Recognition of the interrogator was better in the lower-stress condition, and some participants even confused the gender of their interrogators or guards. Morgan said the findings suggest that a large number of healthy people may not be able to correctly identify people associated with highly stressful events.

The participants were healthy adults in a training setting, not crime victims, so the study cannot capture every feature of a real offence. But it shows that intense stress can undermine recognition even of someone a person has dealt with face to face.

What is weapon focus?

Weapon focus is the tendency for memory of other parts of an event to suffer when a weapon is present. Research usually measures it as poorer memory for the details that appear alongside the weapon, such as the face or clothing of the person holding it.

Two meta-analyses have examined it:

Steblay (1992) reviewed 19 tests and found that weapon presence reduced identification accuracy. The effect was small for line-up identification (.13) and moderate for the accuracy of feature descriptions (.55).
Fawcett, Russell, Peace and Christie (2013) found an overall effect of weapon presence (g = 0.53). It was significantly influenced by the retention interval, how long the witness was exposed to the event, and the level of threat. It did not depend on whether the event took place in a laboratory, a simulation or the real world.

Why does a weapon draw attention away from faces?

Fawcett and colleagues describe two main explanations, which are not mutually exclusive:

The arousal or threat hypothesis. Physiological or emotional arousal narrows attention, so it locks onto the source of danger.
The unusual item hypothesis. A weapon is unexpected in most everyday scenes, and processing something that does not fit takes attention away from everything else.

Their finding that threat influenced the size of the effect is relevant to the first explanation. Either way, the result for testimony is similar: a witness may give a sharp description of the weapon and a vague or mistaken description of the person.

Does a stressful event feel more memorable than it is?

Often, yes. A frightening event can feel intensely vivid, and people naturally assume that a vivid memory is an accurate one. Yale's summary of the Morgan study noted that the research contradicts the assumption that vivid, threatening encounters guarantee reliable identification.

Confidence needs careful handling too. In their 2017 review, Wixted and Wells concluded that a witness's confidence can be informative about accuracy when it is recorded at an initial, uncontaminated test under fair procedures. When those conditions are not met, even a high-confidence identification can be seriously compromised. Confidence that builds over time, through feedback or repeated questioning, is a much weaker guide. For more on how later interviews can shape memory, see how repeated questioning can affect memory.

Stress is only one reason memory can mislead. Why memory is not a perfect recording covers the broader picture.

Does stress during questioning mean someone is lying?

No. Stress affects the person being interviewed, not only their memory of the original event. Being questioned by police about a serious crime is stressful for innocent and guilty people alike, and for witnesses as well as suspects.

In their 2003 meta-analysis of cues to deception, DePaulo and colleagues found that liars appeared somewhat more tense on average. But many behaviours showed no link, or only weak links, to deceit. A small average difference across many studies does not let anyone read an individual. On average, people judge truth and lies correctly only about 54% of the time (Bond and DePaulo, 2006).

There are many innocent explanations for visible stress in an interview:

fear of not being believed;
the memory of a frightening event;
the unfamiliar, formal setting;
worry about unrelated personal matters coming out;
tiredness, grief or shock.

This is why stress responses are best treated as a prompt for better questions, not as evidence. See why there is no single reliable sign of deception and what behavioural leakage is for more.

What does this mean for how testimony is weighed?

The research points in one direction: the conditions under which something was seen matter as much as how confident the witness sounds. Useful questions include:

How stressful was the event, and was a weapon involved?
How long did the witness have to see the person, and how long ago was it?
Was confidence recorded at the first identification, before any feedback?
Is there independent evidence that supports or contradicts the account?

None of these makes a stressed witness worthless. They help explain why an honest, certain witness can still be wrong, which is part of what makes a statement unreliable.

Is there a game that treats stress as a clue, not a verdict?

Murder Book: The Art of Interrogation is an upcoming narrative interrogation game from Gungrounds, an independent studio in Zagreb, Croatia, inspired by real police footage and psychological interviews. Development screenshots show an in-game casebook with an entry called Stress Response, alongside behaviours such as Avoidance and Emotional Fading.

The research above is a good lens for that kind of design. A suspect's stress response might mean something, or it might mean the person is frightened, exhausted or afraid of not being believed. Behaviour is a reason to ask a better question, not proof of guilt, and development screenshots show a case select screen that includes a "not guilty" result.

Murder Book is in development for PC (Steam) and mobile, with release currently planned for the end of Q2 2027. Read What Is Murder Book? for an overview.