See no lies, hear no lies: Differences in discrimination accuracy and response bias when watching or listening to police suspect interviews

AbstractIn the present experiment, police officers attempted to detect truths and lies told by suspects in their police interviews in three different ways: They either saw the suspects (visual condition), only heard the suspects (audio condition) or both saw and heard the suspects (control condition). Research has demonstrated that vocal and speech‐related cues are better diagnostic cues to deceit than visual cues. Therefore, we predicted that participants in the visual condition would perform worst in the lie detection task. Having access only to visual cues may encourage observers to be more reliant on stereotypical beliefs when attempting to detect truths and lies. Since these stereotypes are related to the behaviour of liars, rather than to the behaviour of truth tellers, we further predicted that being exposed only to visual cues may result in a lie bias. The findings supported these hypotheses, and the implications are discussed. Copyright © 2007 John Wiley & Sons, Ltd.

Between-subjects experimental design with 99 British police officers judging 16 video clips from authentic police suspect interviews across three conditions: visual-only, audio-only, and audio-visual control. Signal detection theory measured both discrimination accuracy and response bias.

This study challenged conventional police training by examining whether visual access improves or impairs lie detection accuracy. Using authentic high-stakes police interviews rather than laboratory deception, 99 officers evaluated suspect veracity under three modality conditions. Results revealed a striking paradox: audio-only presentation produced 60% accuracy while audiovisual presentation declined to 47%, supporting the hypothesis that misleading nonverbal cues interfere with judgment. The visual-only condition generated the poorest performance and a pronounced lie bias. These findings suggest that officers' reliance on stereotypical nonverbal behaviors (gaze aversion, fidgeting) systematically increases false positive errors and reduces discrimination ability.

This research has significant operational relevance for police training and interview protocols, suggesting that reducing reliance on visual cues and prioritizing verbal content analysis could substantially improve deception detection accuracy in criminal investigations. The findings call for fundamental restructuring of police interviewing practices and examiner training to emphasize verbal cues over visual stereotypes.

Background & Context

This 2008 experiment emerged from a critical paradox in police deception detection: officers routinely rely on nonverbal visual cues—gaze aversion, fidgeting, nervous movements—yet decades of research demonstrate these stereotypical beliefs are unreliable indicators of deceit. The research revealed that many verbal cues are more diagnostic cues to deceit than nonverbal cues.

The study addressed a fundamental question about presentation modality in lie detection. Police officers watched real life videotaped police-suspect interviews or listened to audiotapes of these interviews, and the officers who listened to the audiotapes were more accurate than those who watched the videotapes. This finding challenged conventional police training and practice.

The research built on earlier work by Mann, Vrij, and colleagues using authentic high-stakes police interviews, representing a significant methodological advancement over laboratory studies with low-stakes deception. The ecological validity of using real suspects in actual criminal investigations made the findings particularly relevant to operational police work.

Research Design & Methodology

In the experiment, police officers attempted to detect truths and lies told by suspects in their police interviews in three different ways: They either saw the suspects (visual condition), only heard the suspects (audio condition) or both saw and heard the suspects (control condition). The study employed a between-subjects experimental design to isolate the effects of presentation modality.

The stimulus materials consisted of 16 video clips from authentic police suspect interviews, drawn from previous research examining high-stakes lies in criminal investigations. Ninety-nine police officers served as judges, evaluating the veracity of suspects across the three experimental conditions. The clips contained an equal number of truths and lies, with ground truth established through independent evidence and case outcomes.

The dependent variables included both discrimination accuracy (correct classification of truths and lies) and response bias, measured using signal detection theory. This dual-measure approach allowed researchers to distinguish between actual discrimination ability and systematic tendencies to judge statements as truthful or deceptive, addressing a known problem in police lie detection research.

Results & Key Findings

The headline finding supported the researchers' prediction: officers in the audio-only condition significantly outperformed those in the visual-only condition. Accuracy in the audio condition (60%) was significantly higher than that in the audio-visual condition (47%), demonstrating that access to visual cues actually impaired deception detection performance.

Key statistical outcomes included:

• Audio-only condition achieved approximately 60% accuracy, significantly above the 47% in the audio-visual condition
• The visual-only condition produced the poorest performance, consistent with the hypothesis that misleading nonverbal cues interfere with accurate judgment
• Being exposed only to visual cues resulted in a lie bias, meaning officers systematically over-attributed deception
• Paying attention to nonverbal cues results in being less accurate in truth/lie discrimination, particularly when only visual nonverbal cues are taken into account, and paying attention to visual nonverbal cues leads to a stronger lie bias

The response bias findings were particularly important for understanding police decision-making. The lie bias observed in the visual condition suggests that reliance on stereotypical nonverbal behaviors (fidgeting, gaze aversion) leads officers to see deception where none exists, increasing false positive errors.

Discussion & Significance

These findings carry profound implications for police training and investigative practice. The authors recommended a change in police practice and argued that for lie detection purposes it may be better to listen carefully to what suspects say. This represents a fundamental challenge to traditional interviewing doctrine emphasizing body language and demeanor.

The research demonstrates why officers trained to watch for stereotypical nonverbal cues perform worse at lie detection. When police officers reported that they watched for fidgeting as a sign of lying, their performance as lie detectors decreased. The vocal and speech-related channel provides more reliable diagnostic information, including speech patterns, response latency, linguistic complexity, and verbal content.

The study contributes to a growing body of evidence suggesting that cognitive-verbal approaches to deception detection outperform behavioral observation methods. By demonstrating this effect with experienced police officers judging real suspects in high-stakes contexts, the research substantially strengthened the ecological validity of the modality effect previously observed primarily in laboratory settings.

Limitations & Considerations

The study used pre-recorded interview clips rather than live interactive interviews, which may limit generalizability to actual investigative contexts where officers conduct their own questioning. The between-subjects design, while methodologically sound, meant individual officers only experienced one condition, preventing within-subjects comparison of their performance across modalities.

The sample consisted of British police officers from a specific geographic region and time period, raising questions about generalizability across different policing cultures and training backgrounds. Additionally, while the interviews represented authentic criminal cases, the specific crimes and suspect characteristics may not represent the full range of deception contexts police encounter.

Ground truth verification, while based on case outcomes and independent evidence, may not be perfect—some suspects classified as deceptive may have been truthful and vice versa, potentially affecting the accuracy benchmarks.

Practical Applications

The findings suggest police training programs should de-emphasize visual behavioral cues and increase focus on verbal content analysis and vocal characteristics. Interview recording and review procedures might benefit from audio-only analysis to reduce the contaminating influence of misleading visual stereotypes. This could be particularly valuable in secondary review of interviews or in training scenarios.

For consumers and defendants, these results raise important questions about the reliability of police credibility assessments based on demeanor evidence. The systematic lie bias observed when officers focus on visual cues suggests potential for false accusations, particularly concerning for vulnerable populations who may display nervous behaviors for reasons unrelated to deception. The research supports arguments for more objective, content-based credibility assessment methods in criminal justice contexts.