Outsmarting the liars: The benefit of asking unanticipated questions

This 2009 study by Vrij, Leal, Granhag, Mann, Fisher, Hillman, and Sperry tested whether asking unanticipated questions could improve deception detection by exploiting liars' inability to prepare for unexpected lines of inquiry. Pairs of liars and truth tellers were interviewed individually about a shared restaurant visit, with the interviewer progressing from predictable opening questions to unexpected spatial and temporal questions and a restaurant sketch task. Results confirmed the hypothesis: up to 80% of liar and truth-teller pairs were correctly classified based on response consistency to unanticipated questions, particularly drawings — far exceeding chance-level performance on anticipated questions.

Laboratory analog study in which pairs of participants (liars who fabricated a shared restaurant visit, truth tellers who genuinely shared the experience) were interviewed individually using a structured protocol progressing from anticipated opening questions to unanticipated spatial questions, temporal questions, and a drawing task; cross-pair response consistency served as the primary veracity indicator.

This laboratory study addressed the fundamental problem that people perform near chance-level (54%) at detecting deception through behavioral observation. Researchers tested whether asking unanticipated questions could exploit liars' reliance on prepared cover stories by having pairs of participants interviewed separately about a shared restaurant experience. The experimental design contrasted responses to anticipated questions versus unanticipated spatial/temporal questions and drawing tasks, measuring cross-pair response consistency as the detection cue. Results showed that while truth-telling pairs drawing on genuine shared memories produced consistent accounts, liar pairs who had independently fabricated stories showed greater inconsistencies when faced with unexpected questions. The approach achieved up to 80% accuracy in correctly classifying deceptive versus truthful pairs, representing a substantial improvement over traditional deception detection methods.

This cognitive load approach offers polygraph examiners a complementary strategy for improving detection accuracy by incorporating unexpected spatial, temporal, and visual questions that exploit liars' preparation limitations. The findings support developing structured interview protocols that systematically include unanticipated elements to maximize differential cognitive demands between truthful and deceptive respondents.

Background & Context

Decades of deception research have consistently revealed a sobering truth: people are remarkably poor lie detectors. Five decades of lie-detection research have shown that people's ability to detect deception by observing behavior and listening to speech is limited — with, on average, only 54% of truths and lies being correctly classified. This near-chance performance level has driven researchers to seek fundamentally new approaches to investigative interviewing.

One critical but underappreciated insight underpins this research: it is possible to ask questions that raise cognitive load more in liars than in truth tellers. This cognitive lie-detection perspective consists of two core approaches — the imposing-cognitive-load approach, which aims to make the interview setting more difficult for interviewees, and the strategic-questioning approach, which examines different ways of questioning that elicit the most differential responses between truth tellers and liars.

A key vulnerability for liars is that their carefully constructed cover stories depend on anticipating what questions will be asked. A consistent finding in deception research is that liars prepare themselves when anticipating an interview. Planning makes lying easier, and planned lies typically contain fewer cues to deceit than spontaneous lies. However, the positive effects of planning will only emerge if liars correctly anticipate which questions will be asked — and investigators can exploit this limitation directly. Vrij, Leal, Granhag, Mann, Fisher, Hillman, and Sperry (2009) set out to test this principle empirically with a carefully designed laboratory study.

Research Design & Methodology

The study hypothesised that the responses of pairs of liars would correspond less with each other than would responses of pairs of truth tellers, but only when those responses were given to unanticipated questions. Liars and truth tellers were interviewed individually about having had lunch together in a restaurant. This elegant scenario allowed the researchers to study deception within a realistic, socially plausible context where pairs of participants shared a common cover story — or a genuine shared memory.

The experimental design leveraged a within-subjects contrast between anticipated and unanticipated question types. Key design elements included:

• Pairs of participants — both liars and truth tellers were interviewed separately, and cross-pair response consistency was used as the primary detection signal.
• Anticipated questions — conventional opening questions such as "What did you do in the restaurant?" which liars could readily prepare for.
• Unanticipated spatial questions — for example, "In relation to the front door and where you sat, where were the closest diners?"
• Unanticipated temporal questions — for example, "In which order did you discuss the different topics you mentioned earlier?"
• Drawing task — participants were also asked to draw the layout of the restaurant, an entirely novel and unanticipated request.

After the interview, participants confirmed which questions had come as a surprise, validating the experimental manipulation of anticipation. The research team assessed cross-pair response consistency as the primary cue to veracity — the key theoretical insight being that truth-telling pairs, drawing on genuine shared episodic memory, would produce more overlapping accounts than liar pairs who had independently fabricated their stories.

Results & Key Findings

The central finding was striking: unanticipated questions enabled correct classification of up to 80% of liar and truth-teller pairs — a result that stands in dramatic contrast to the near-chance accuracy achieved with standard, anticipated questions.

The results supported the hypothesis, and based on correspondence in responses to the unanticipated questions, up to 80% of liars and truth tellers could be correctly classified, particularly when assessing drawings. This represents a substantial improvement over baseline lie-detection accuracy and over what was achievable using anticipated questions in the same study.

• Anticipated (opening) questions: Pairs of liars were not classified above chance level based on response overlap to anticipated questions — liars' prepared stories aligned sufficiently to evade detection.
• Unanticipated spatial questions: Liar pairs showed significantly less cross-pair consistency than truth-teller pairs, as liars independently constructed incompatible accounts of spatial details they had never actually experienced.
• Unanticipated temporal questions: A similar pattern emerged — liars independently fabricated timelines that diverged, while truth tellers recalled a shared sequence of actual events.
• Drawing task: The drawing condition produced particularly strong discrimination, with the highest classification accuracy among all unanticipated question types.
• Verbal vs. drawn responses: Truth tellers' and liars' drawings differed more from each other than did their verbal recalls, highlighting the drawing request as an especially powerful elicitation format.

The theoretical mechanism behind these results is rooted in memory and fabrication asymmetry. Truth tellers encode events along multiple sensory and contextual dimensions — spatial layouts, temporal sequences, sensory impressions — providing a rich, consistent episodic foundation. Liars, constructing cover stories in advance, tend to focus on plausible narrative content while neglecting the fine-grained peripheral details that unanticipated questions probe.

Discussion & Significance

This research represents a meaningful paradigm shift in how deception detection is conceptualised. Rather than searching for inherent behavioral or physiological "tells" that liars inadvertently emit — an approach with a weak empirical track record — the unanticipated questions technique actively engineers conditions under which deception becomes detectable. It demonstrates that it is possible to ask questions that raise cognitive load more in liars than in truth tellers, and that this strategic-questioning approach can elicit the most differential responses between truth tellers and liars.

The drawing finding is particularly theoretically rich. Truth tellers' and liars' drawings differed more from each other than did their verbal recalls, suggesting that liars were inclined to omit critical details — partly because those details did not exist in their experience, and partly to avoid triggering further questions about elements they could not credibly account for. This "masking strategy" is significantly harder to execute when forced to spatially represent an environment in a sketch.

There is evidence that asking unanticipated questions is beneficial to deception detection regarding past events — including mock crimes — extending the applicability of this technique beyond the specific restaurant scenario tested here. The study sits within a broader and growing research programme demonstrating that cognitive and strategic interviewing methods outperform both traditional behavioral cue detection and physiological methods when detection accuracy is the primary goal.

Limitations & Considerations

Several important constraints on these findings should be acknowledged:

• Laboratory analog design: The study used a controlled mock scenario, which limits direct generalisation to high-stakes real-world criminal or national security interviews where motivations, consequences, and deception sophistication may differ substantially.
• Pair-based detection logic: The primary classification method relied on cross-pair consistency, meaning two suspects must be interviewed independently and their responses compared. This approach is not always operationally feasible in field settings.
• Drawing coding novelty: The research team noted that since drawing coding had never been done before in this context, reliability codings were obtained for all drawings rather than just a selection. Further, the level of correspondence between drawings and the actual restaurant could not be measured for liars, since it was unknown whether they drew an existing restaurant or which restaurant that might have been.
• Anticipation verification: The unanticipated nature of questions was verified post-interview, which is methodologically sound but means the technique's effectiveness depends on maintaining genuine surprise — a condition that could erode as awareness of this approach spreads.
• Follow-up replication concerns: Subsequent research has provided more mixed results, with some studies finding that unanticipated questions failed to improve veracity judgment accuracy above chance for interviewers in live interview settings.

Practical Applications

For investigators and interviewers, this research offers a clear and actionable prescription: do not rely solely on predictable, anticipated lines of questioning. Incorporating spatially specific and temporally specific questions — particularly requests to sketch environments, reconstruct timelines in non-obvious formats, or describe peripheral sensory details — can meaningfully improve the ability to detect collaborative deception. The technique is especially well-suited to situations where multiple suspects are interviewed separately, as the cross-pair consistency analysis provides a powerful signal that is difficult for coordinated liars to fake without shared episodic experience.

The findings also carry important implications for polygraph and forensic interview practice more broadly. Unanticipated questioning is, in principle, countermeasure-resistant — liars cannot prepare answers to questions they do not anticipate. Examiners who integrate strategic, unanticipated question formats into pre-test interviews or structured investigative interviews may be able to elicit verbal and behavioral cues that complement physiological assessments, producing a more complete and reliable picture of veracity.