Cues to deception and ability to detect lies as a function of police interview styles

In Experiment 1, we examined whether three interview styles used by the police, accusatory, information-gathering and behaviour analysis, reveal verbal cues to deceit, measured with the Criteria-Based Content Analysis (CBCA) and Reality Monitoring (RM) methods. A total of 120 mock suspects told the truth or lied about a staged event and were interviewed by a police officer employing one of these three interview styles. The results showed that accusatory interviews, which typically result in suspects making short denials, contained the fewest verbal cues to deceit. Moreover, RM distinguished between truth tellers and liars better than CBCA. Finally, manual RM coding resulted in more verbal cues to deception than automatic coding of the RM criteria utilising the Linguistic Inquiry and Word Count (LIWC) software programme. In Experiment 2, we examined the effects of the three police interview styles on the ability to detect deception. Sixty-eight police officers watched some of the videotaped interviews of Experiment 1 and made veracity and confidence judgements. Accuracy scores did not differ between the three interview styles; however, watching accusatory interviews resulted in more false accusations (accusing truth tellers of lying) than watching information-gathering interviews. Furthermore, only in accusatory interviews, judgements of mendacity were associated with higher confidence. We discuss the possible danger of conducting accusatory interviews.

Two-experiment design with 120 mock suspects in staged event paradigm interviewed using three different styles (accusatory, information-gathering, behaviour analysis), followed by 68 police officers judging videotaped interview clips. Verbal content analyzed using CBCA and Reality Monitoring with both manual and automated (LIWC) coding methods.

This landmark study challenged the effectiveness of confrontational interrogation by systematically comparing three police interview styles for their ability to elicit verbal deception cues. Using mock suspect interviews analyzed with both manual and automated coding methods, the research demonstrated that accusatory approaches—characteristic of the Reid Technique—produced fewer diagnostic verbal indicators than information-gathering or behaviour analysis methods. The study further revealed that manual coding of Reality Monitoring criteria detected more deception cues than automated linguistic software (LIWC), suggesting that human expert analysis captures contextual nuances missed by automated systems. These findings have significant implications for interrogation practice, supporting the shift toward rapport-based, information-gathering interview approaches and highlighting limitations of fully automated deception detection.

The findings provide empirical support for moving away from confrontational Reid Technique-style interrogation toward information-gathering and structured behaviour analysis approaches in police interviews. The superiority of manual coding over automated analysis underscores the continued importance of human expert judgment in verbal deception detection, while suggesting that any fully automated system must be validated against human expert performance on contextually sensitive verbal cues.

Background & Context

This landmark 2007 study by Aldert Vrij, Samantha Mann, Susanne Kristen, and Ronald Fisher addresses a critical question facing law enforcement worldwide: which interview style produces the most reliable verbal cues to deception? The research emerged during a pivotal period when many Western countries were transitioning from confrontational, accusatory interrogation methods toward more information-gathering approaches like the PEACE model.

The study directly challenged the traditional Reid Technique's accusatory approach, which dominated American policing and assumed interviewers could detect deception through confrontation. By systematically comparing three distinct interview styles—accusatory, information-gathering, and behaviour analysis—Vrij and colleagues provided empirical evidence about which methods actually elicit diagnostic verbal content. This research also represented one of the first rigorous comparisons of manual versus automated coding of verbal deception cues, testing whether computer programs like LIWC could match human expert analysis.

The work builds on decades of verbal deception detection research using Criteria-Based Content Analysis (CBCA) and Reality Monitoring (RM), two structured content analysis methods originally developed in European forensic contexts. Understanding how interview style affects the production and detection of these verbal cues has profound implications for both investigative practice and wrongful conviction prevention.

Research Design & Methodology

The study comprised two experiments. In Experiment 1, 120 mock suspects either told the truth or lied about a staged event and were interviewed by a police officer using one of three interview styles. Participants were randomly assigned to truth-teller or liar conditions. Truth-tellers participated in an actual staged event, while liars were given scripted information about the event they had not actually experienced.

The three interview styles represented distinct policing approaches:

• Accusatory interviews: Confrontational questioning typical of the Reid Technique, where the interviewer presumes guilt and challenges the suspect's denials
• Information-gathering interviews: Open-ended, rapport-based questioning focused on eliciting detailed narratives (similar to PEACE model)
• Behaviour Analysis Interview: Systematic questioning protocol examining behavioral patterns and response consistency

Interview transcripts were coded using CBCA and Reality Monitoring methods. The study compared manual RM coding with automatic coding using the Linguistic Inquiry and Word Count (LIWC) software programme. In Experiment 2, 68 police officers watched videotaped interviews from Experiment 1 and made veracity and confidence judgements.

Results & Key Findings

The results showed that accusatory interviews, which typically result in suspects making short denials, contained the fewest verbal cues to deceit. This finding directly challenges the effectiveness of confrontational interrogation methods for eliciting diagnostic verbal content.

Key findings from Experiment 1 (verbal cues):

• Reality Monitoring distinguished between truth tellers and liars better than CBCA
• Manual RM coding resulted in more verbal cues to deception than automatic coding of the RM criteria utilising the LIWC software programme
• Information-gathering and behaviour analysis interviews elicited significantly more diagnostic verbal content than accusatory approaches
• The superior performance of manual coding suggests that automated linguistic analysis may miss subtle contextual cues that human coders detect

Key findings from Experiment 2 (detection accuracy):

• Accuracy scores did not differ between the three interview styles; however, watching accusatory interviews resulted in more false accusations (accusing truth tellers of lying) than watching information-gathering interviews
• Only in accusatory interviews, judgements of mendacity were associated with higher confidence
• This dangerous combination—higher false accusation rates paired with greater confidence—creates particular risk in accusatory contexts

Discussion & Significance

The findings provide strong empirical support for information-gathering interview approaches over accusatory methods. The authors discuss the possible danger of conducting accusatory interviews—a danger that manifests both in reduced verbal cue production and increased false accusations of innocent suspects. When accusatory interviews prompt suspects to give brief denials rather than detailed narratives, investigators lose access to the very verbal content that could reveal deception.

The Reality Monitoring superiority over CBCA represents an important methodological finding. RM's inclusion of "cognitive operations" as a lie indicator (liars tend to describe more thought processes and reasoning) provides a bidirectional cue system, whereas CBCA criteria primarily indicate truthfulness. This makes RM more diagnostically powerful in real-world applications where establishing both truthfulness and deception is essential.

The confidence-accuracy relationship in accusatory interviews is particularly troubling. When observers watch accusatory interviews and judge someone as lying, they feel more confident in that judgment—yet their accuracy doesn't improve. This overconfidence effect, combined with increased false accusations, suggests that accusatory methods may create a dangerous illusion of investigative certainty. This has clear implications for false confession research and wrongful conviction prevention.

Limitations & Considerations

The study employed a mock-crime laboratory paradigm with university participants, which limits ecological validity compared to actual criminal investigations with genuine stakes. Real suspects face imprisonment, not experimental penalties, which fundamentally changes motivation and stress. The staged event (playing Connect 4) lacks the emotional intensity and complexity of actual crimes.

The police officer participants in Experiment 2 viewed edited video clips rather than conducting live interviews themselves, which removes the interactive dynamics of real interrogation. Officers' judgments may differ when they control questioning strategy versus passively observing. Additionally, the study examined only initial interviews; real investigations involve multiple interview sessions where strategies may evolve.

The manual coding of CBCA and RM criteria requires extensive training and remains somewhat subjective despite standardized protocols. Different coders may reach different conclusions about the same verbal content, though the study employed reliability checks.

Practical Applications

Law enforcement agencies should recognize that accusatory interview techniques are counterproductive for both eliciting verbal cues and avoiding false accusations. The evidence supports transitioning to information-gathering approaches that encourage suspects to provide detailed narratives. Investigators should employ open-ended questions that allow suspects to talk extensively, as verbal content—not behavioral observations—provides more reliable deception cues.

For polygraph examiners and deception detection professionals, the findings emphasize the importance of interview technique in pre-test and post-test phases. Rather than confronting examinees with accusations, examiners should use narrative-eliciting questions that produce analyzable verbal content. The RM criteria, particularly attention to cognitive operations and sensory details, offer practical frameworks for statement analysis that complement physiological measurement.

Police trainers should emphasize Reality Monitoring criteria over CBCA given its superior diagnostic performance, and invest in training for manual coding rather than relying solely on automated linguistic analysis. Most importantly, training should address the dangerous overconfidence that accusatory methods produce, helping investigators recognize when their confidence exceeds their actual accuracy.