Police officers ability to detect deception in high stakes situations and in repeated lie detection tests

Thirty-seven police officers, not identified in previous research as belonging to groups that are superior in lie detection, attempted to detect truths and lies told by suspects during their videotaped police interviews. In order to measure consistency in their ability, the officers each participated in four different tests, each of which was on a different day. They were asked to indicate their confidence in being able to distinguish between truths and lies prior to the first test and after completing all four tests. We predicted that accuracy rates would be higher than those typically found in research with police officers; that good or poor performances on an individual test would be partly caused by luck, and, consequently, participants' accuracy scores were likely to progress towards the mean if their performance on all four tests was to be combined; and that officers would underestimate their own performance. These hypotheses were supported. Copyright © 2006 John Wiley & Sons, Ltd.

Background & Context

This 2006 study addresses a critical gap in deception detection research: the longstanding reliance on low-stakes laboratory studies using student participants. Prior research typically found police officers detecting deception at only around 55-60% accuracy—barely better than chance, leading many to conclude that professional lie-catchers possess no special expertise.

However, consistent findings from high-stakes research suggested that police officers may utilize reliable cues to deception when the stakes are elevated. The study builds on earlier work by Vrij, Mann, and colleagues showing that ecological validity matters: real-world police interviews involving serious crimes produce different behavioral cues than laboratory experiments. This research innovatively examines not just whether officers can detect deception, but whether their performance is consistent across repeated tests—a critical but understudied question.

Research Design & Methodology

Thirty-seven police officers participated in four separate deception detection tests, each conducted on a different day, viewing videotaped suspect interviews where ground truth was known. The officers were specifically selected because they had not been previously identified as belonging to groups with superior lie detection ability—making them representative of typical law enforcement personnel.

Key methodological features included:

• High-stakes stimulus materials: Real police interviews with actual suspects in serious criminal cases
• Repeated testing design: Four tests across different days to assess consistency and control for luck
• Confidence ratings: Officers indicated their confidence before the first test and after completing all four tests
• Within-subjects design: Each officer served as their own control across the four testing sessions

The researchers predicted that accuracy would be higher than typical police officer studies; that individual test performance would be partly due to luck and thus regress toward the mean across all four tests; and that officers would underestimate their own performance.

Results & Key Findings

Officers achieved a total accuracy of 72% when performance across all four tests was combined—a finding substantially above the chance level of 50% and well above the typical 55-60% seen in most deception studies with police.

Key statistical findings included:

• Officers detected truths with 70% accuracy and lies with 73% accuracy, showing balanced performance without a truth bias
• Performance on individual tests showed greater variability, supporting the prediction that luck influenced single-test results
• Combining results across all four tests demonstrated regression toward the mean, validating that repeated testing provides more reliable estimates
• Officers were "overly modest" about their performance, believing they had only performed at chance level

All three hypotheses were supported, confirming that high-stakes contexts improve detection accuracy, that single-test performance is unreliable, and that officers lack accurate insight into their own capabilities.

Discussion & Significance

The findings challenge the prevailing narrative that police officers are no better than laypeople at detecting deception. The 72% accuracy represents a substantial improvement over the 50-60% typically found, suggesting police officers may not be as poor at deception detection as commonly believed. However, this improvement appears contingent on the nature of the lies being detected.

High-stakes situations differ fundamentally from laboratory-based low-stakes lies, and police officers appear to utilize reliable cues to deception specifically in these high-stakes contexts. The repeated-measures design revealed an important methodological insight: single assessments can be misleading due to random variation. The regression toward the mean observed across four tests suggests that extraordinary performances (very high or very low) on individual tests often reflect luck rather than stable ability.

The disconnect between officers' actual performance and their self-assessment raises important questions about training and professional confidence. Officers' lack of awareness of their competence may stem from receiving little objective feedback in real-world settings, where ground truth often remains uncertain.

Limitations & Considerations

The study acknowledges several constraints on generalizability. Officers did not conduct the interviews themselves but watched recordings made by others, which removes interactive elements that might aid or hinder real-world detection. The sample size of 37 officers, while adequate for within-subjects analysis, limits broader conclusions.

The stimulus materials, though ecologically valid, came from a specific context—British police interviews with suspects in serious crimes. Performance might differ with different interview styles, cultural contexts, or types of deception. The study also cannot specify which behavioral cues officers used to achieve their accuracy, leaving open questions about the mechanism underlying their improved performance.

Finally, while 72% accuracy is substantially better than chance, it still means officers made errors on more than one-quarter of judgments—a non-trivial error rate with potentially serious consequences in criminal justice contexts.

Practical Applications

For law enforcement practice, these findings suggest that officers' intuitive judgments about suspect veracity in high-stakes interviews may have more validity than laboratory research would suggest. However, the tendency toward excessive modesty about detection abilities highlights a training gap: officers need better calibration between their actual performance and their confidence.

The repeated-testing methodology offers lessons for evaluating deception detection skills in training contexts. Single assessments should be interpreted cautiously, as they may reflect situational factors or chance rather than stable ability. For consumers of polygraph and investigative services, the research underscores that context matters enormously—conclusions from low-stakes laboratory studies may not apply to real criminal investigations, where motivational intensity and consequences fundamentally alter behavioral cues.