A recent study published in the Journal of Investigative Psychology and Offender Profiling has revealed the impact of stress and biased instructions on eyewitness accuracy. Researchers found that participants who experienced a high-stress, simulated hostage scenario had low rates of correct identifications during lineup tasks conducted a week later. Biased instructions during the lineup further decreased overall accuracy, increasing the likelihood of choosing incorrectly. The findings underscore the challenges of eyewitness identifications in high-stress situations and the potential for flawed procedures to exacerbate errors.
Eyewitness testimony plays a significant role in criminal investigations, but it is also a leading cause of wrongful convictions. Despite extensive research on eyewitness accuracy, most studies have relied on controlled, artificial scenarios, such as videos or photos. These laboratory conditions often fail to replicate the heightened stress and complexity of real-world events, such as violent crimes. Stress, in particular, is thought to influence memory, with conflicting theories suggesting it could either impair or enhance recall, depending on the situation.
The new study sought to address this gap by using a more realistic and stressful setting: a simulated hostage scenario. The researchers aimed to examine how such conditions influence the accuracy of eyewitness identifications, as well as the effects of biased versus unbiased lineup instructions. By doing so, they hoped to provide insights that could help improve the reliability of eyewitness identifications in real-life cases.
“Eyewitness identification plays a crucial role in the criminal justice system, but it can be influenced by mistakes and biases,” said study author Thomas J. Nyman, a lecturer in forensic psychology at the University of Reading. “We wanted to explore how a simulated hostage situation, designed to replicate a highly stressful, violent, and threatening scenario, affects eyewitness identification accuracy.”
The researchers worked with 122 participants from an international criminal justice training program, conducted by the United Nations. “By working with the United Nations in Italy, we had a unique chance to conduct this research in a setting that closely mimics real-life situations, which is something traditional laboratory studies cannot do,” Nyman explained. “This collaboration gave us the opportunity to study eyewitness decision-making in ways that are not usually possible in standard academic research.”
Participants, who were from diverse backgrounds, underwent a high-stress, 45-minute hostage simulation. The scenario included mock insurgents armed with realistic weapons, physical intimidation, and a staged abduction. The setup aimed to replicate the chaotic and threatening conditions of a real-life violent crime.
One week after the event, participants completed multiple lineup tasks designed to identify the insurgents. Each lineup included three photos and was presented on a computer screen. Some lineups contained a photo of an actual perpetrator (“target present”), while others did not (“target absent”). To test the influence of bias, participants were randomly assigned to receive either biased or unbiased instructions.
Unbiased instructions informed participants that the perpetrator might not be in the lineup and encouraged them not to guess. In contrast, biased instructions implied that the perpetrator was present and urged participants to make an effort to choose someone, even if unsure.
During the task, the researchers recorded participants’ choices, their confidence in their decisions, and the time they took to respond. Altogether, the study analyzed 1,030 lineup decisions, providing a comprehensive dataset for understanding how stress and instruction type influenced accuracy.
For lineups with the perpetrator present, participants identified the correct individual only 38% of the time. For lineups where the perpetrator was absent, participants correctly rejected the lineup 54% of the time.
“We were surprised by the level of accuracy participants achieved despite the challenges,” Nyman told PsyPost. “That does not mean that accuracy was high, but the study involved high stress, multiple perpetrators, and a one-week delay between the event and the line-up. These factors are typically expected to lower accuracy substantially.”
Biased instructions had a significant impact, decreasing overall accuracy by promoting more guessing and incorrect choices. When biased instructions were used, participants were more likely to make a selection in both target-present and target-absent lineups. This increased the hit rate (correctly identifying a perpetrator when present) but also significantly raised the false alarm rate (incorrectly identifying someone in target-absent lineups). Specifically, false alarms were twice as common under biased instructions compared to unbiased ones.
The study also found a link between confidence and accuracy. Participants who expressed high confidence in their choices were more likely to be correct, particularly when unbiased instructions were given. However, biased instructions inflated confidence in incorrect choices, weakening the relationship between confidence and accuracy.
Response times provided additional insights. Participants who quickly rejected target-absent lineups were more likely to be correct. However, response time did not reliably predict accuracy in target-present lineups, suggesting that confidence and decision-making processes differ depending on the type of lineup.
“Our study shows that how a line-up is presented affects accuracy,” Nyman explained. “Biased instructions can help catch the guilty person but also increase the risk of mistakes, while unbiased instructions reduce errors but may miss the guilty person. Conducted in one of the most realistic hostage simulations to date, our findings are directly applicable to real-world settings, showing that biased line-up instructions lead to decreased accuracy.”
While the study provides valuable insights, it also has limitations. The use of a three-person lineup differs from the six- or eight-person lineups commonly used in real investigations, potentially inflating accuracy rates. Additionally, the researchers could not control for all variables that might influence identification, such as lighting conditions or the “weapon focus effect,” where attention is drawn to a weapon rather than the perpetrator’s face.
“While the realism of our study is a major strength, it also introduced challenges, such as the smaller line-up size and no control group,” Nyman noted. “Additionally, we could not isolate the effects of stress and multiple perpetrators. However, our findings align with existing research, suggesting these limitations did not undermine the core insights.”
The researchers recommend further studies to address these limitations and to explore other factors affecting eyewitness accuracy, such as race and cultural biases, varying levels of stress, and lineup procedures. They also emphasize the need for more ecologically valid research to better understand how real-world conditions impact memory and identification.
“Our long-term goal is to understand the factors that affect a person’s memory in real-life situations, such as stress, visual distance, lighting, and own-race bias,” Nyman said. “By examining how these factors impact eyewitness accuracy, we aim to help the justice system improve how eyewitness identifications are conducted and interpreted.”
“This study represents a rare intersection of academic research and international collaboration. Partnering with the United Nations and the Italian Army allowed us to conduct the most realistic investigation to date of eyewitness identification accuracy following a simulated hostile event. These insights are invaluable for informing practices that could help mitigate injustice worldwide.”
The study, “Eyewitness identifications based on biased or unbiased line-up instructions after a realistic and violent hostage simulation,” was authored by Thomas J. Nyman, Giulia Cappa, Angelo Zappalà, and Pekka Santtila.
