Did somebody see it? Applying the verifiability approach to insurance claims interviews

This 2014 study by Nahari, Leal, Vrij, Warmelink, and Vernham was the first to apply the Verifiability Approach (VA) to insurance claim interviews, examining whether truth tellers and liars differ in their provision of verifiable details and their reporting of witnesses. Using 83 true and false statements about damage, theft, or loss, the researchers found that the witness dimension — not the verifiable detail count — distinguished liars from truth tellers, with 86% of liars reporting unwitnessed incidents compared to 54% of truth tellers.

An analog laboratory study coding 83 true and false insurance claim statements (related to damage, theft, or loss) along two dimensions: witness presence and the number of perceptual and contextual details that could be independently verified by an investigator.

This study extended the Verifiability Approach from police interrogation contexts to insurance claim interviews, addressing the massive problem of insurance fraud detection. Researchers analyzed 83 true and false insurance claim statements related to damage, theft, and loss, coding them for witness presence and verifiable details. The key finding revealed an asymmetry: witness presence successfully differentiated truth tellers from liars, but the core VA prediction about verifiable details did not hold in this insurance context. This represented the first application of the VA to insurance investigations, conducted by researchers from Bar-Ilan University and University of Portsmouth.

The findings suggest that witness-based indicators may be more reliable than detail verifiability in insurance fraud detection, requiring adaptation of the VA framework for different investigative contexts beyond traditional polygraph applications.

Background & Context

Insurance fraud represents a massive and largely invisible financial problem across the global economy. Deception research regarding insurance claims is rare but relevant given the financial loss in terms of fraud. Despite the scale of the problem, investigators have historically lacked evidence-based, scientifically validated tools for identifying fraudulent claimants through verbal statement analysis. This study stepped into that gap, extending a promising new paradigm — the Verifiability Approach (VA) — into the insurance domain for the first time.

The Verifiability Approach was originally developed for police interrogation contexts, grounded in the intuition that honest and dishonest accounts differ in a fundamental strategic way. The verifiability approach states that truth tellers and liars differ from each other in terms of the number of details they give that can be verified. Put simply, truthful people willingly include checkable specifics because they have nothing to hide, while liars deliberately avoid them to reduce their exposure. Truth tellers tend to be verbally forthcoming, freely disclosing checkable details they can remember.

By 2014, the VA had demonstrated encouraging results in alibi contexts, but had not yet been tested in the distinctive environment of insurance claim interviews — where liars face a unique strategic challenge. Lie detection in insurance claim settings is difficult as liars can easily incorporate deceptive statements within descriptions of otherwise truthful events. Nahari and colleagues recognized this challenge and designed a study specifically to probe whether the VA's core predictions would hold up under real-world insurance-claim conditions.

Research Design & Methodology

The study was authored by Galit Nahari (Bar-Ilan University, Israel) and co-authored by Sharon Leal, Aldert Vrij, Lara Warmelink, and Zarah Vernham (University of Portsmouth, UK), combining expertise from two leading research centers in investigative psychology. Eighty-three true and false insurance claim statements, related to damage, theft, or loss, were coded in terms of 'witnesses' (was the incident witnessed by others) and 'verifiability' (the number of perceptual and contextual details provided that could be checked by the investigator).

This design was distinctive in using two separate coding dimensions rather than just one, allowing the researchers to separately evaluate whether the presence of witnesses and the density of verifiable details each contributed to distinguishing truth from deception. The insurance claim scenarios were ecologically grounded — covering the everyday categories of damage, theft, and loss — lending the findings relevance to real investigative practice. The approach of having participants report on genuine recent experiences or fabricate plausible incidents is consistent with established laboratory analog methodology in deception research.

Key design elements included:

• Sample: 83 true and false insurance claim statements
• Statement types: Damage, theft, and loss claims
• Coding dimension 1 — Witnesses: Whether the incident was reported as having been witnessed by others
• Coding dimension 2 — Verifiability: Number of perceptual and contextual details that could be independently checked by an investigator
• Domain: Insurance claim interview setting (a novel context for the VA at the time)
• Affiliated institutions: Bar-Ilan University (Israel) and University of Portsmouth (UK)

Results & Key Findings

The headline finding was a striking asymmetry: the witness dimension successfully differentiated liars from truth tellers, but the verifiability-of-details dimension did not. This pattern was unexpected given the VA's prior success in other settings and introduced important nuance to the theory.

Nahari, Leal, Vrij, Warmelink, and Vernham (2014) found that liars were much more likely to report that there were no witnesses to the incident (86%) than truth tellers (54%). This large gap in witness reporting was the study's most actionable finding — fraudulent claimants overwhelmingly staged or described incidents that had no observers, while truthful claimants were roughly split on the question.

Key findings at a glance:

• 86% of liars described incidents with no witnesses, compared to 54% of truth tellers
• The majority of liars, compared with half the truth tellers, described unwitnessed incidents — and this difference between the groups allowed for the detection of liars only.
• Discrimination between liars and truth tellers based on the verifiability of details was not possible.
• The VA predicts that liars solve this dilemma by selectively leaving out details that investigators can verify, and including more details that are difficult to verify. In the insurance context, this second prediction was not supported.
• The witness factor — not the detail verifiability count — emerged as the diagnostic signal in this specific domain

The null result on verifiable detail counts is itself a theoretically significant outcome. In fact, subsequent meta-analysis included this study among approximately 21% of studies that obtained null findings on the verifiable details dimension — making it a pivotal calibration point for the broader VA literature.

Discussion & Significance

The study's most important contribution is its identification of a domain-specific mechanism by which liars manage their deception in insurance contexts. Rather than simply providing fewer verifiable details — the classic VA prediction — insurance fraudsters appear to strategically construct claims around unwitnessed events. The authors reasoned that liars used embedding strategies, which allowed them to report sufficient checkable detail to appear credible due to the context in which the lie is told. By setting their fabrications in solitary contexts, liars sidestep the need to manage verifiable detail counts at all.

This finding highlighted an important limitation of applying VA in its unmodified form to insurance settings. Subsequent research directly built upon this insight. Notifying claimants about the interviewer's intention to check their statements for verifiable details increased accuracy rates from around chance level to around 80%. When an additional model statement was introduced alongside the information protocol, the accuracy climbed further. If the information protocol is additionally supplemented with a model statement, the VA even attains a discrimination rate of 90%.

This study also holds significance within the broader arc of VA research by demonstrating that the approach is context-sensitive. The applicability of the verifiability approach, originally developed for assessing the veracity of suspect accounts within police interrogation settings, differs across judgmental settings, and these differences influence application. Nahari et al. (2014) were the first to map one such difference in a rigorous, peer-reviewed study.

Limitations & Considerations

Several methodological constraints temper the study's conclusions. The use of a laboratory analog paradigm — even one grounded in realistic scenarios — cannot fully replicate the motivational stakes of a genuine insurance fraud investigation. Real fraudsters face financial gain or loss and may therefore show different patterns of strategic information management than experimental participants.

• Ecological validity: Analog insurance scenarios may not capture the full complexity of genuine fraud cases, where claimants often have prolonged preparation time
• No information protocol: The VA was tested without the "information protocol" component; later research showed this omission likely suppressed the verifiable-details effect
• Detection asymmetry: The witness dimension enabled detection of liars only — it did not provide a positive signal for truth tellers, limiting its bilateral diagnostic utility
• Sample size: With 83 statements total, the study is relatively small, which limits statistical power to detect moderate effects on the verifiability-of-details dimension
• Null result on core VA measure: The absence of a difference in verifiable detail counts raises questions about the boundary conditions of the VA that later research has continued to address

Practical Applications

For insurance fraud investigators and claims managers, the study's witness dimension finding offers a practical screening heuristic: claimants who report incidents that were entirely unwitnessed warrant heightened scrutiny. While no single verbal cue is definitive on its own, the dramatic gap between liar and truth-teller rates of witness-free reporting — 86% versus 54% — suggests that routine documentation of witness presence or absence during claims interviews could improve fraud detection. The VA posits that truthful statements should contain more verifiable details than deceptive ones, so specifically asking interviewees to provide verifiable details and informing them that their credibility will be assessed based on the number of verifiable details they provide — an instruction called the 'Information Protocol' — should maximize the difference in verifiable details between truthful and deceptive statements.

More broadly, the study underscores that verbal credibility assessment tools must be tailored to the investigative context in which they are used. What appears to be required to improve lie detection accuracy is employing proactive interview protocols that impose manipulations to elicit and magnify cues to deception. This paper was instrumental in motivating the development of those enhanced protocols — making it a foundational reference for anyone designing structured credibility assessment procedures in the insurance, legal, or investigative psychology domains.