Detecting Deception: Why Lie Detection Is Hard
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Detecting deception is an inference about whether a person intentionally communicates something they believe to be false or deliberately creates a misleading impression. It is difficult because the observable signs of speaking, remembering, worrying, thinking, and concealing information overlap. In a criminal investigation, a confident judgment that someone "looks dishonest" can change the direction of questioning, affect the handling of evidence, and expose an innocent person to suspicion. The scientifically defensible approach examines claims, context, independently established facts, interview conditions, and the limitations of every assessment method.
A widely cited meta-analysis of 206 research documents involving 24,483 judges found an average accuracy of approximately 54% when people classified truthful and deceptive communications without special assistance. The same research reported about 47% correct identification of lies and 61% correct identification of truths. These are averages across particular study designs, not the accuracy of every investigator or interview and not a universal probability that any given speaker is lying. Bond and DePaulo's meta-analysis establishes the starting point: spontaneous human lie detection is considerably less dependable than confidence in one's intuition often suggests.
What Does Detecting Deception Mean?
Deception involves communication intended to produce or maintain a belief that the communicator considers false or misleading. A lie is a central form of deception, but communication can mislead through omission, selective presentation, staged evidence, or technically accurate statements arranged to support a false conclusion. Researchers must specify which behaviors count as deception before measuring detection accuracy. A statement that is factually inaccurate can originate in honest error, memory distortion, misunderstanding, intoxication, fatigue, poor communication, or lack of information.
This distinction is especially consequential in forensic psychology. The psychological construct of deceptive intent concerns a person's knowledge and communicative purpose. The factual truth of an event must be established from evidence. A witness's confidence is a behavioral observation, not an independent verification of accuracy. A physiological measurement or questionnaire score is an assessment result. An investigator's interpretation is an opinion. Whether statements satisfy a legal definition of fraud, perjury, obstruction, or false reporting is a matter of applicable law and adjudicated facts. These categories answer different questions and must remain separate.
Why Is Lie Detection So Hard?
Truthful and deceptive people often behave similarly
The central obstacle is weak separation between the distributions of behavior shown by truthful and deceptive speakers. Someone telling the truth may be frightened, evasive about embarrassing details, hesitant, unusually formal, or inconsistent about peripheral events. Someone lying may be relaxed, articulate, rehearsed, cooperative, and persuasive. The familiar image of a nervous liar confuses arousal with deception. Research synthesized in DePaulo and colleagues' review of 158 proposed cues found that many candidate behaviors had no discernible association with lying or only weak associations.
The same behavior has multiple possible causes
A long pause may reflect deliberate invention, but it may also reflect memory search, language translation, uncertainty, fear of giving a misleading answer, cognitive disability, or a poorly understood question. Avoided eye contact may reflect social convention, personal comfort, or a person's reaction to authority. Changes in speech rate can arise from emotion, unfamiliar surroundings, or illness. A meaningful cue must discriminate between these competing explanations in the relevant population, rather than merely appear plausible to an observer.
Observers bring expectations into the judgment
People do not watch behavior as neutral measuring instruments. They interpret what they notice using stereotypes, prior knowledge, emotional reactions, and their hypotheses about a case. Once an interviewer suspects a lie, ambiguous behavior can be interpreted as confirmation. That judgment may then shape subsequent questions and the interviewee's reactions, creating a feedback loop. The concept of truth-default theory, introduced by Timothy Levine, also highlights how ordinary communication is built around an inclination to accept what others say until a reason for doubt emerges. Both excessive trust and excessive suspicion can produce errors when substituted for corroboration.
Stakes, motivation, and preparation vary
The psychological conditions of a casual social lie differ from those of a person responding to an allegation that could change their life. Motivation may affect impression management, emotional arousal, and attempts to control behavior. Preparation affects fluency. Truth tellers can rehearse; liars can improvise. No single cognitive or emotional pathway uniquely identifies deliberate falsehood. Laboratory findings need to be interpreted against the stakes, opportunity for preparation, interview setting, and definition of ground truth in each study.
What Does Research Say About Lie Detection Accuracy?
The 54% figure is a useful benchmark for unaided human judgments in the studies reviewed by Bond and DePaulo. It is close to the accuracy expected from guessing when the test material contains roughly equal numbers of truths and lies. It should not be presented as a universal law of every deception task. Some interviewing methods have produced higher performance in controlled comparisons, and the proportion of lies in real-world settings can differ radically from the balanced samples often used in experiments.
Accuracy also conceals two distinct kinds of error. A false positive occurs when a truthful person is judged deceptive. A false negative occurs when a deceptive person is judged truthful. Sensitivity refers to the proportion of deceptive communications correctly detected; specificity refers to the proportion of truthful communications correctly accepted. Positive predictive value asks a different question: among statements labeled deceptive, how many are actually deceptive? That proportion depends on how common deception is in the population and context being assessed.
Imagine, purely as an illustration, that 100 statements contain 10 deceptive ones and that a hypothetical method detects 80% of lies while correctly clearing 80% of truths. It would flag 8 actual lies and 18 truthful statements, yielding 26 alerts, of which only 8 would be correct. The positive predictive value would be approximately 31%. This arithmetic is an example of base-rate effects; it does not describe the measured performance of a specific lie detector. In low-prevalence screening, an apparently impressive test can produce many more accusations than verified deceptions.
Do police officers and other professionals perform better?
Professional experience does not automatically confer reliable access to a person's private intentions. Meta-analytic evidence does not support assuming that investigators, judges, clinicians, or security officers can identify lies accurately merely because they frequently encounter high-stakes communication. Performance depends on the specific task, information available, quality of independent evidence, and method of questioning. The person who knows a case file may detect contradictory facts that a blind observer cannot; that is an informational advantage rather than evidence of a special ability to read a face. Bond and DePaulo provide the general benchmark, while Vrij, Hartwig, and Granhag's review examines why behavior-based confidence so often exceeds diagnostic accuracy.
Can You Tell Someone Is Lying by Body Language?
Averted gaze, crossed arms, facial touching, fidgeting, shifting posture, looking upward, and supposedly meaningful eye movements are popular signals in online lie-detection guides. None provides a scientifically validated standalone rule for identifying a lie. Body movement may be informative about discomfort, attention, interactional dynamics, or emotional activation, but a link between discomfort and intentional falsehood must be demonstrated rather than assumed.
Microexpressions are very brief facial movements sometimes treated as involuntary disclosures of hidden truth. A facial expression can convey emotion without specifying its cause, and an emotion is not identical to deception. Anger may arise from an unfair accusation; fear may arise from a threatening environment; amusement may arise from an unrelated thought. The Annual Review of Psychology synthesis characterizes discovered nonverbal deception cues as faint and unreliable and explains why popular stereotypes persist despite contrary evidence.
An observer's belief that a particular gesture "always means lying" is especially dangerous when interviews concern traumatized witnesses, autistic people, people with communication disabilities, speakers using an additional language, or anyone operating under unfamiliar social expectations. Variation in presentation should initiate careful clarification and accommodation, not an automatic inference about guilt or credibility.
Are There Reliable Verbal Signs of Lying?
Verbal information can be more useful than gestures when it is elicited through a sound interview and compared with independent facts. Relevant features can include the specificity of verifiable details, plausible chronology, a speaker's ability to explain what they directly perceived, and consistency with evidence that was acquired independently. Yet no particular word, pronoun, pause, grammatical construction, or amount of detail is proof of deception.
Researchers have examined approaches such as criteria-based content analysis, statement-validity assessment, reality monitoring, and linguistic classification. Their evidence, validation populations, and proposed uses differ; an observed average separation between truthful and deceptive accounts does not establish forensic accuracy for a single witness. Life experience, education, language competence, trauma, memory quality, coaching, narrative style, and the questions asked can all alter what appears in an account.
A truthful witness may omit detail because they never encoded it, because time has passed, or because repeated interviews have changed retrieval conditions. A deceptive speaker may provide richly detailed but unverifiable narrative material. Investigators therefore gain more from looking for evidence that could independently confirm or disconfirm a claim than from scoring how vividly a story is told.
Does Cognitive Load Reveal a Lie?
Some theories predict that fabricating, monitoring, and maintaining a lie may impose additional demands on working memory and executive control. Investigative interviewing research has used unexpected questions, requests for fuller accounts, and procedures intended to make differences in available knowledge easier to observe. These are research-based approaches, but their effects depend on the task and the interviewee. A truth teller who is anxious, sleep-deprived, neurodivergent, or remembering a complex event may also struggle with cognitively demanding questions.
A meta-analysis by Vrij, Fisher, and Blank examined cognitive approaches that included increasing cognitive load, encouraging fuller reporting, and asking unexpected questions. In the subset directly comparing approaches, the reviewed studies reported better combined classification accuracy than traditional approaches. The paper describes approximately 71% total accuracy under the cognitive approach versus 56% under the comparison approach, with different truth- and lie-detection results. These numbers are study-level comparisons, not proof that the procedures will achieve those rates in every real investigation.
Increasing difficulty merely to make a person stumble is an especially weak application of cognitive theory. Questions should be relevant and understandable; reasonable language, disability, and developmental accommodations must be preserved. The objective of a defensible interview is to obtain accurate information, not to manufacture stress and then interpret it as a symptom of deception.
How Does the Strategic Use of Evidence Work?
Strategic use of evidence, commonly abbreviated SUE in the research literature, examines the relationship between an interviewee's account and independently collected information. Investigators can first obtain an open account and then explore discrepancies using carefully controlled disclosure of evidence. The central question becomes whether a claim fits verifiable facts, rather than whether the speaker displays nervousness.
In an experimental study of police trainees, Hartwig and colleagues found that training in strategic evidence use changed how interviewers questioned mock suspects and improved deception judgments under the study conditions. The published report included an accuracy comparison of 85.4% versus 56.1% in its particular design. Such findings support testing structured interviewing methods while also requiring replication, independent validation, and care in translating laboratory or mock-crime outcomes to actual casework.
Evidence inconsistency is an investigative lead rather than automatic proof of lying. Records can be incomplete, mistaken, mislabeled, or collected under an incorrect assumption. A person may not have known the relevant information, may misunderstand a date, or may have been exposed to contaminating details through the investigation itself. Any contradiction should be tested against alternate factual explanations.
Can a Baseline Reveal Sudden Dishonesty?
"Baselining" commonly means learning how a person normally behaves and treating later departures from that pattern as suspicious. A change from customary speech or gesture might indicate that something in the interaction has changed, but identifying its cause is the harder inference. A shift in the topic, perceived threat, interviewer behavior, fatigue, cultural norms, or emotional salience can produce deviations without intentional falsehood.
The value of comparing different statements depends on what is being compared and on the quality of the interview design. A 2024 study in Acta Psychologica examined baselining within structured interviewing protocols, illustrating that baselining is an empirical question about a specified method, not a universal trick for identifying deception. A person's "normal" response during casual conversation is not necessarily their normal response during an accusatory interview.
Is a Polygraph a Lie Detector?
The polygraph records physiological responses, typically including cardiovascular, respiratory, and electrodermal activity, while an examiner uses a particular questioning procedure. These measurements reflect physiological activity rather than a direct sensor of whether a statement is factually false or intentionally misleading. Interpretation depends on comparison questions, response patterns, examiner decisions, and the testing context.
The National Research Council's consensus report concluded that research quality, setting, and purpose matter greatly to polygraph validity. Estimates from certain specific-incident studies cannot simply be carried over to broad personnel or security screening. Countermeasures, variability between individuals, false positives, and base rates complicate results. Polygraph claims also require separate legal analysis because admissibility and permitted use differ among jurisdictions. The technique deserves its own technical evaluation; this article addresses only why it should not be mistaken for direct observation of deception.
Can Brain Scans, Voice Analysis, or Eye Tracking Detect Lies?
Brain activity and autonomic signals can correlate with psychological processes involved in particular deception tasks, but a brain image, vocal stress trace, pupil response, or eye movement does not inherently encode the legal truth of a statement. Laboratory classification tasks often train systems on known instructed lies, which can differ from remembered events, spontaneous omissions, disputed allegations, and complex intentions. Small datasets, overfitting, unrepresentative participants, and changes in equipment or interview conditions make generalization uncertain.
Technical performance needs external validation on the intended population, independently established truth labels, blinded assessment, and reporting of sensitivity, specificity, calibration, uncertainty, and failures. Promotional claims of very high accuracy are insufficient without those foundations. The National Research Council reviews the theoretical and evidentiary complications of physiological and alternative lie-detection approaches. A scientific measurement of an associated process remains distinct from proof of guilt or a legal finding.
Can Artificial Intelligence Detect Deception More Accurately?
Machine-learning systems may classify text, voice, images, or multimodal recordings using features learned from training datasets. Their outputs can look precise even when the learned distinction depends on features specific to a laboratory dataset, background conditions, presentation style, or recording quality. A model trained on acted lies may fail on spontaneous interviews; one trained on native speakers may misclassify speakers using a second language. Cross-validation within a dataset cannot replace testing on independent people and genuinely different settings.
In a 2025 Journal of Communication investigation, Markowitz and Levine reported 12 studies of AI-generated veracity judgments using one large language model under varying tasks, media, and truth–lie proportions. Performance and bias varied substantially across contexts, and the authors urged caution about using certain large language models for deception detection. This is informative emerging evidence, not a comprehensive verdict on every possible AI architecture.
If an automated system labels someone "deceptive," a responsible evaluation asks what outcome was predicted, how ground truth was established, which populations were tested, how often truthful people were flagged, whether accuracy survives shifts in prevalence, and whether users can contest errors. The numerical confidence of software is not a forensic conclusion. AI-generated explanations may sound compelling even when the model's classification lacks a reliable evidentiary basis.
Why Do False Positives Matter So Much in Investigations?
The practical costs of a false positive depend on what happens after a label. An erroneous suspicion can influence whom investigators question next, how intensely they interrogate, whether exculpatory material is pursued, and how later reviewers interpret a recorded interaction. Treating apparent nervousness as proof of lying can convert ordinary distress into confirmation of an initial hypothesis. The resulting process may appear coherent while progressively reducing the consideration of alternatives.
A valid investigation must preserve the distinction between the probability of deception, the strength of evidence for a particular factual claim, and the criminal law's requirements for responsibility. No deception screening score determines guilt. In the United States, legal rules on expert opinions and admissibility concern methodology and fit to the case; they do not transform a behavioral observation into proof. The federal courts' discussion of the Rule 702 amendment emphasizes the proponent's burden to establish admissibility requirements for expert evidence. Other jurisdictions apply their own evidentiary rules.
Lie Detection, Credibility, Reliability, and Factual Accuracy
These terms frequently blur together in popular discussions, but forensic work benefits from precise use. Deception concerns intentionally misleading communication. Factual accuracy concerns whether an assertion corresponds to what happened. Reliability concerns the dependability of a source or method under defined conditions. Credibility is a broader judgment of whether a source deserves belief, affected by context and potentially by legally specified criteria. A truthful witness may be inaccurate; a deceptive witness may include many accurate details.
A confident witness can sincerely believe a memory that has changed over time. An evasive witness may be protecting private information unrelated to the offense. A suspect may falsely confess, producing a statement that sounds self-incriminating despite being untrue. Accordingly, the assessment of eyewitness identification, false confessions, forensic interviewing, and deception detection intersects, but each has a separate evidentiary literature and a separate canonical search intent.
How Interviewing Can Improve the Quality of Information
A defensible investigative interview begins with clear objectives and an open opportunity to explain events in the person's own words. Interviewers should distinguish what the person perceived directly from what they heard from others, what they inferred, and what they learned after the event. Relevant follow-up questions should be intelligible, non-coercive, and designed to test information without contaminating the account.
The United Nations' 2024 Manual on Investigative Interviewing for Criminal Investigation promotes systematic, rapport-based, non-coercive information gathering. The Méndez Principles likewise describe professional safeguards, respect for rights, and interviewing techniques aimed at reliable information. These documents provide authoritative guidance; their recommended procedures and safeguards must still be implemented under the law governing a particular investigation.
Recordings help other professionals examine question wording, interviewer behavior, disclosures of evidence, breaks, and whether alleged contradictions arose from the interview itself. The APA's 2022 resolution on criminal suspect interrogations recommends complete recording and recognizes the risks associated with coercive practices, misleading evidence presentations, and false confessions. It is a professional policy statement grounded in research, not a claim that a particular technique eliminates all errors.
Can Pressuring a Suspect Make a Lie Easier to Detect?
Pressure may make behavior more conspicuous while making interpretation less valid. It can increase stress for innocent and guilty people alike and may damage memory retrieval, communication, or willingness to cooperate. When an interviewer assumes guilt in advance, a person's denial can be reframed as further evidence of dishonesty, even though denial is also the natural response of an innocent suspect.
The objective of information-gathering interviewing is to test alternative explanations and obtain accounts that can be checked. False confessions demonstrate a particularly consequential failure of equating a persuasive or self-incriminating statement with factual accuracy. The APA's 2022 interrogation resolution addresses interrogation duration, recording, vulnerability, fabricated evidence, minimization, and related safeguards. Specific rules about police questioning, access to counsel, admissibility, and remedies are established by applicable law and vary by jurisdiction.
Cultural, Developmental, and Clinical Fairness
Behavioral norms vary across cultures, communities, interpersonal settings, and languages. Interviewers must be especially cautious when interpreters are used or a person has communication differences. Pauses, speech rhythm, gaze, politeness, expressiveness, directness, and narrative organization can be altered by language proficiency and social context. A method that has not been validated across relevant groups can create systematic errors while appearing neutral.
Psychiatric diagnoses, autism, ADHD, intellectual disability, trauma history, and neurological conditions should not be treated as evidence that a person is deceptive. These may be relevant to how questions are understood, how information is recalled, and which procedural accommodations are needed. Equally, a diagnosis does not determine whether someone's account is true. Investigators should consider comprehension, accessible language, interpretation, fatigue, vulnerability, and the possibility that an interview setting itself creates distress.
The Difference Between Laboratory Research and Real Cases
Controlled studies can define ground truth, randomize participants, and isolate methods. That experimental control is valuable for estimating causal effects of an interviewing procedure. Yet instructed or incentivized deception, mock crimes, short video clips, homogeneous participant groups, and an artificially balanced ratio of truths and lies may not reproduce the social and legal complexity of an actual investigation.
Field research can better approximate real stakes, but often faces incomplete ground truth, case-selection bias, changing interview tactics, selective recording, and limited access to exculpatory information. An investigator should therefore ask whether a method was evaluated with reliable truth criteria; whether judges were blinded; whether data from the same person leaked into training and testing; whether the intervention itself changed interviewee behavior; and whether the findings were replicated across institutions.
A model or procedure can perform well in one laboratory and poorly in another. Performance must also be considered as an operating characteristic of a complete system, including training, question design, the interpreter of results, and the consequences attached to the classification.
Calibration, Uncertainty, and the Legal Meaning of an Assessment
An accuracy rate summarizes correct classifications at a particular threshold. It does not show whether the system's confidence is calibrated. A calibrated assessment that assigns an 80% probability to a class should be correct roughly 80% of the time in comparable, independently assessed cases; such calibration must be tested rather than inferred from a model's output. In a forensic context, analysts should report limitations, potentially different error rates among subgroups, the relevant population, and uncertainty about applying study results to an individual.
There is no universal legal definition of lying that a psychological device can directly measure. False statement offenses, perjury, admissibility of evidence, credibility instructions, and procedural protections are defined by jurisdiction-specific law. An expert may explain what research says about memory, stress, interviewing, and behavioral cues, but a psychological opinion and a legal finding occupy different roles. The evidence must remain open to adversarial examination and independent factual testing.
A More Reliable Way to Evaluate a Suspicious Statement
A good evidence-centered process identifies exactly what factual proposition is disputed, what independent evidence can test it, and which alternative explanations would also fit the observed behavior. Preserve original records and the full interview context. Ask open questions before introducing suggestive information. Distinguish a genuine contradiction from a detail the person never claimed to know. Consider the effects of repeated questioning, prior information exposure, misunderstanding, translation, and poor retrieval conditions.
An unsupported conclusion such as "the speaker touched their face, so the denial was a lie" should be replaced by a documented account of what was said, which facts were externally verified, which remain unresolved, and how the questions were asked. If an interview reveals new testable facts, assess them independently. Do not let a single behavior or algorithmic score substitute for the investigative work of verification.
In everyday relationships, the same reasoning argues for questions, context, patterns over time, and independently checkable facts rather than amateur interrogation. Research on police interviews cannot simply be exported into romantic disputes, workplace conflicts, or family conversations. Coercive amateur lie detection can harm trust while producing little reliable information.
Common Myths About Detecting Deception
Myth: Liars always avoid eye contact
Eye contact is shaped by culture, social expectations, attention, and the conditions of the interaction. It has not emerged as a reliable standalone deception marker in major evidence syntheses. Vrij and colleagues document the weakness of nonverbal detection cues.
Myth: A person who changes their story is obviously lying
A genuine contradiction is relevant to investigate, especially if it concerns a central fact, but recall can change with time, questioning, exposure to other accounts, and correction of prior mistakes. The interviewer must determine what each version actually claims and whether independent evidence resolves the discrepancy.
Myth: Nervous people are more likely to be dishonest
Nervousness is an emotional or physiological response with many possible causes. Fear of being wrongly accused may produce the same or greater distress as fear of being caught lying. Treating arousal as a lie indicator creates a predictable route to false accusations.
Myth: Experienced interrogators can read microexpressions as proof
Expertise in conducting interviews and gathering evidence is valuable. Expertise does not turn expressions into direct measurements of deceptive intent. The scientific literature calls for evidence-based judgments and acknowledgment of the faint and unreliable nature of many proposed nonverbal cues.
Myth: A machine's percentage makes the answer objective
A polygraph, voice analyzer, or AI model converts data into an output under assumptions. Errors in the training sample, measurement, ground truth, interpretation, threshold, or population can bias results. Mathematical precision in an output is not the same as validity or legal proof.
Frequently Asked Questions
What are the most reliable signs that someone is lying?
No single facial expression, gesture, vocal change, or phrase reliably establishes lying. Independently verified contradictions and information elicited through carefully designed questions can be more informative, while alternative explanations for discrepancies still need examination. The evidentiary value comes from corroboration and context.
How accurate are people at detecting lies?
Bond and DePaulo's large meta-analysis found approximately 54% correct classifications across its reviewed unaided judgments, including about 47% of lies and 61% of truths correctly identified. Performance varies across settings, tasks, interview methods, and truth–lie base rates. The figure is a historical research benchmark, not a fixed rate for each individual.
Can facial expressions or body language prove deception?
No. Facial movements, gaze, posture, and nervous gestures can reflect many processes and usually show weak, inconsistent relationships with intentional falsehood. They can guide respectful clarification but should not establish an accusation.
Are police or judges better at spotting lies?
Occupation and exposure alone do not guarantee strong accuracy. Domain knowledge and access to independently verified evidence may improve factual evaluation, but intuition about demeanor remains limited. Any specialized technique needs task-relevant evidence of performance.
Can a lie detector test tell if someone is telling the truth?
Polygraphs measure physiological responses; they do not directly record lying. Accuracy estimates depend on technique, purpose, population, and research quality. Results must be understood through false-positive and false-negative rates and the jurisdiction's rules on use or admissibility.
Can AI analyze video and accurately detect lies?
Some AI systems can classify experimental recordings, but current research has documented task sensitivity and serious biases. A 2025 study of one language model found sharply different outcomes across conditions. Independent real-world validation is essential before applying any such tool to a person's legal interests.
Does remembering something incorrectly count as lying?
An honest memory error is different from intentional deception. Determining whether someone was deliberately misleading requires evidence about knowledge and intent, which cannot be read directly from an inaccurate statement or a nervous reaction.
What is the difference between deception detection and interrogation?
Deception detection is a research and practical question about identifying intentional misleading communication. Investigative interviewing is a process for obtaining and checking information. Interrogation practices, including their legal authorization and safeguards, vary by jurisdiction. Information gathering can be effective without treating the interview as a contest to produce a confession.
Do inconsistencies mean a witness is unreliable?
An inconsistency calls for evaluation of its importance, origin, and relation to independent evidence. Peripheral memory details may vary without changing the account's central accuracy. A deliberate falsehood, a genuine misrecollection, and an ambiguous question require different interpretations.
Is there a scientifically perfect lie detector?
No method discussed here has established error-free identification of deception across people, contexts, and real legal cases. Methods should be evaluated on validated performance, uncertainty, population fit, safeguards, and whether they produce independently testable evidence.
For broader context, see Investigative Psychology: What It Is and How It Differs From Criminal Profiling.
Related Articles
References
American Psychological Association. (2022). Resolution on interrogations of criminal suspects. Official APA policy.
Bogaard, G., van de Bovekamp, A. M., & Colwell, K. (2024). Assessing the efficacy of baselining in deception detection: A comparative analysis of the reality interview and structured interview. Acta Psychologica, 242, 104112. https://doi.org/10.1016/j.actpsy.2023.104112.
Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and Social Psychology Review, 10(3), 214–234. https://doi.org/10.1207/s15327957pspr1003_2.
DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129(1), 74–118. https://doi.org/10.1037/0033-2909.129.1.74.
Hartwig, M., Granhag, P. A., Strömwall, L. A., & Kronkvist, O. (2006). Strategic use of evidence during police interviews: When training to detect deception works. Law and Human Behavior, 30, 603–619. https://doi.org/10.1007/s10979-006-9053-9.
Levine, T. R. (2014). Truth-default theory (TDT): A theory of human deception and deception detection. Journal of Language and Social Psychology, 33(4), 378–392. https://doi.org/10.1177/0261927X14535916.
Markowitz, D. M., & Levine, T. R. (2025). The (in)efficacy of AI personas in deception detection experiments. Journal of Communication, 75(6), 459–469. https://doi.org/10.1093/joc/jqaf034.
Méndez Principles Steering Committee of Experts. (2021). Principles on Effective Interviewing for Investigations and Information Gathering. Official principles.
National Research Council. (2003). The Polygraph and Lie Detection. National Academies Press. https://doi.org/10.17226/10420.
U.S. Courts. (2023). Recent and proposed amendments to federal rules: Amendments to Federal Rule of Evidence 702. U.S. Courts annual report.
United Nations. (2024). Manual on Investigative Interviewing for Criminal Investigation. UN Human Rights Knowledge Gateway.
Vrij, A., Fisher, R. P., & Blank, H. (2017). A cognitive approach to lie detection: A meta-analysis. Legal and Criminological Psychology, 22(1), 1–21. https://doi.org/10.1111/lcrp.12088.
Vrij, A., Hartwig, M., & Granhag, P. A. (2019). Reading lies: Nonverbal communication and deception. Annual Review of Psychology, 70, 295–317. https://doi.org/10.1146/annurev-psych-010418-103135.