Abstract
Lie detection, which the Medical Subject Headings vocabulary classifies under criminal psychology, is the attempt to distinguish truthful from deceptive statements by measuring physiological, behavioural, verbal, or neural signals. The polygraph’s Comparison Question Test assumes that lying produces a distinctive arousal state; the Concealed Information Test instead asks whether an examinee recognises details only a guilty person would know; while other work scores behavioural cues or images the brain. The sobering finding of a century of research is that no cue is a reliable signature of deception, and even the strongest methods perform far worse in the field than their advocates claim. This article defines the field, maps its methods, and examines why low base rates make screening treacherous, with three interactive demonstrations of the base-rate problem, the Concealed Information Test, and the weak link between cues and accuracy.
Keywords: lie detection, polygraph, deception detection, cues to deception, base rate
Lie detection is the effort to determine, from measurable signs rather than corroborating evidence, whether a person is telling the truth. The oldest and best-known instrument is the polygraph, which records autonomic activity — heart rate, blood pressure, respiration, and electrodermal response — on the assumption that deception is accompanied by arousal that a trained examiner can interpret (National Research Council, 2003). The premise is intuitive and old, but it conceals a problem that has shadowed the field since its beginning: there is no known physiological response unique to lying. Every signal a polygraph records can be produced by fear, surprise, indignation, or the simple stress of being suspected, so the instrument measures arousal, not deception, and the inference from one to the other is where the difficulty lies.
That gap has driven the field in two directions. One accepts that no lie-specific signal exists and asks a different question — not “is this person aroused?” but “does this person recognise information only the culprit would know?” — the logic of the Concealed Information Test (Lykken, 1959). The other abandons instruments for the speaker’s own behaviour, cataloguing the facial, vocal, and verbal cues that supposedly betray a liar (Ekman & Friesen, 1969). Both have been studied for decades, and both have been humbled by the same discovery: the cues are weak, the base rates are unforgiving, and human judges perform barely above chance (Bond & DePaulo, 2006). The sections below define the field, map its methods, and weigh the evidence honestly.
Key Takeaways
- Lie detection infers deception from physiological, behavioural, verbal, or neural signals; no signal is a unique signature of lying, so every method infers deception indirectly (National Research Council, 2003).
- The polygraph’s Comparison Question Test reads autonomic arousal, but arousal has many causes, and countermeasures can defeat it (Honts et al., 1994).
- The Concealed Information Test asks whether an examinee recognises guilty knowledge rather than whether they are aroused, giving it a controllable and much lower false-positive rate (Ben-Shakhar & Elaad, 2003).
- Behavioural and verbal cues to deception are weak on average, and untrained observers judge truth and lies at about 54% accuracy — barely above chance (DePaulo et al., 2003; Bond & DePaulo, 2006).
- Low base rates make screening applications treacherous: even an accurate test flags far more innocent people than guilty ones when deception is rare (National Research Council, 2003).
What Lie Detection Is
In the Medical Subject Headings vocabulary, lie detection is defined as the procedure for identifying deception by analysis of physiological responses to a graduated series of questions — a definition that reflects the historical dominance of the polygraph but understates how varied the field has become. What unites its methods is the goal, not the signal: each tries to separate truthful from deceptive accounts using something other than independent corroboration. Where they differ is in what they measure and, more deeply, in the inference they make from that measurement.
The foundational obstacle is that deception is a mental act with no dedicated physiology. Anger raises blood pressure; fear quickens the pulse; concentration slows breathing. A liar may show all of these, but so may an innocent person who is frightened of being disbelieved, and a practised liar may show none of them. This is the reason a century of research has failed to find a “Pinocchio’s nose” — a behavioural or bodily response that occurs when and only when someone lies (Luke, 2019). Every operational method is therefore built on a proxy, and the quality of a method is the quality of its proxy.
Those proxies fall into four families. The autonomic approach, embodied in the polygraph, reads bodily arousal. The memory-based approach, embodied in the Concealed Information Test, reads recognition of crime-relevant details. The behavioural approach reads the speaker’s face, voice, and words. The neural approach, the youngest, reads brain activity during deceptive responding with event-related potentials or functional imaging (Farwell & Donchin, 1991; Langleben et al., 2002). Each family rests on a different theory of what deception does to a person, and each inherits a characteristic failure mode, examined in turn below.
The four families of lie-detection method
| Family | Signal measured | Core assumption | Characteristic weakness |
|---|---|---|---|
| Autonomic (polygraph) | Heart rate, respiration, electrodermal response | Lying raises arousal an examiner can read | Arousal has many causes; defeated by countermeasures |
| Memory-based (Concealed Information Test) | Orienting response to recognised crime details | Only the guilty recognise concealed details | Needs withheld details; detects recognition, not guilt |
| Behavioural | Facial, vocal, and verbal cues | Deception leaks through poorly controlled behaviour | Cues are weak; judges perform near chance |
| Neural | P300 event-related potential; fMRI activity | Recognition or effort has a brain signature | Poor individual classification; countermeasure-prone |
The Polygraph and the Comparison Question Test
The polygraph does not detect lies; it records physiological activity while an examiner asks questions and then interprets the pattern. The dominant protocol in field use is the Comparison Question Test, which contrasts a subject’s reactions to relevant questions (a direct probe of the act under investigation, such as whether the examinee took the money) with reactions to comparison questions designed to be arousing for everyone (a broad probe such as whether the examinee had ever, before age 25, taken something that was not theirs). The theory holds that the guilty react more strongly to the relevant questions and the innocent more strongly to the comparison questions (National Research Council, 2003).
The most authoritative assessment of this method, the U.S. National Research Council’s 2003 review, reached a measured but damning conclusion. The polygraph performs better than chance in controlled laboratory studies of specific-incident testing, but its accuracy is well short of what its use in high-stakes decisions would require, and it rests on a theory of arousal that lacks scientific support. The review was especially critical of security screening, where the base rate of deception is very low, warning that in such settings a polygraph programme would either miss most spies or falsely implicate a great many loyal employees (National Research Council, 2003). The scientific community shares this scepticism: surveys of psychophysiologists and of the broader scientific community found that most doubt the theoretical rationale of the Comparison Question Test and would not accept polygraph evidence as reliable (Iacono & Lykken, 1997).
A further weakness is that the test can be beaten. Because it depends on producing a larger reaction to comparison than to relevant questions, an examinee who covertly augments their responses to the comparison questions — by mental arithmetic, controlled breathing, or a discreet muscle contraction — can defeat it. Laboratory work shows that both physical and mental countermeasures, taught in a brief training session, substantially reduce the accuracy of polygraph tests, and that examiners often cannot detect their use (Honts et al., 1994).
The first demonstration makes the screening problem concrete: adjusting a test’s accuracy and the base rate of deception shows how many of the people it flags are actually innocent.
The base-rate problem: who fails the test?
Set a test’s accuracy and the rate of deception in 10,000 people. The bar shows, of everyone the test flags, how many are actually deceptive (navy) versus innocent (gold).
| Test: fail | Test: pass | |
|---|---|---|
| Actually guilty (100) | 80 | 20 |
| Actually innocent (9,900) | 1,980 | 7,920 |
With the default 80% accuracy and a 1% base rate, fewer than 4% of those who fail are actually deceptive. This is why the National Research Council judged the polygraph unsuitable for low-prevalence screening: the rare true positives are swamped by false ones.
The Concealed Information Test
The Concealed Information Test — historically the Guilty Knowledge Test — abandons the search for a lie-specific reaction and asks a different question entirely. Rather than probing whether a person is lying, it probes whether they recognise information that only the perpetrator of a crime would possess. For each item, the examinee is presented with the crime-relevant detail (the actual murder weapon, say) among several plausible alternatives, and their orienting response — typically electrodermal, but also respiratory and cardiac — is compared across alternatives. A guilty examinee, recognising the true detail, tends to react most strongly to it; an innocent examinee, unable to tell it from the decoys, reacts at random (Lykken, 1959).
The great advantage of this design is that its error rate is calculable and controllable. An innocent examinee has, on any single item with m equally plausible alternatives, only a one-in-m chance of reacting most strongly to the critical detail by coincidence. Across several independent items, the probability of an innocent person appearing guilty on all of them falls steeply, so the false-positive rate can be pushed as low as the number of items allows. A meta-analytic review confirms that the test discriminates well, with especially strong protection of the innocent: false-positive rates are low, and accuracy is highest when multiple questions and objective electrodermal scoring are used (Ben-Shakhar & Elaad, 2003).
The Concealed Information Test is not without limits. It can only be used when investigators possess salient details that were withheld from the public and that a guilty suspect would necessarily have encoded, which restricts its applicability; and it detects recognition, not guilt, so an innocent person who learned the details through the media or interrogation can fail it. A broad appraisal of behavioural, autonomic, and neural deception measures concludes that the Concealed Information Test rests on the firmest theoretical footing of the physiological methods, precisely because it measures memory rather than the will-o’-the-wisp of arousal, while still counselling modesty about field performance (Meijer et al., 2016).
The second demonstration shows why the test protects the innocent so well: increasing the number of questions or the number of alternatives per question drives the chance of a false positive toward zero.
Why the Concealed Information Test protects the innocent
An innocent examinee cannot tell the true detail from the decoys, so on each item they react to the critical one only by chance. Add items or alternatives and the odds of failing them all collapse.
Unlike the polygraph, whose error rate rests on a contested theory of arousal, the Concealed Information Test’s false-positive rate is a matter of arithmetic the examiner controls by choosing how many items and alternatives to use.
Behavioural and Verbal Cues
Long before instruments, people believed they could catch a liar by watching and listening — a shifting gaze, a nervous laugh, a story too smooth or too ragged. The scientific study of this intuition began with the idea of nonverbal leakage: the proposal that concealed emotions escape through channels a liar controls poorly, such as the feet and the fleeting expressions of the face (Ekman & Friesen, 1969). The hope was that deception, being effortful and emotionally charged, would leave a behavioural residue an observer could learn to read.
Decades of accumulation have largely disappointed that hope. The definitive meta-analysis of cues to deception examined 158 candidate behaviours and found that most bear no reliable relation to deception at all, and that even the strongest cues have small effect sizes; liars are, if anything, slightly more tense and give accounts that are less compelling and less forthcoming, but no behaviour approaches the reliability of a diagnostic sign (DePaulo et al., 2003). A recent reappraisal argues that even these modest effects are probably overstated, inflated by small samples and publication bias, so that the true cues to deception are weaker still than the canonical review suggests (Luke, 2019).
The consequence for human judges is stark. A meta-analysis of hundreds of studies found that people distinguish truths from lies at an average accuracy of about 54% — barely above the 50% expected by chance — and that neither professional experience nor confidence improves this much (Bond & DePaulo, 2006). The most productive modern response has been to stop treating deception as a passive emotional leak and to make lying cognitively harder instead: because maintaining a fabrication while under scrutiny is mentally taxing, interview techniques that increase cognitive load — asking for detail, requesting the account in reverse order, posing unanticipated questions — can widen the small behavioural gap between liars and truth-tellers (Vrij et al., 2010; Vrij et al., 2019).
Detection accuracy across methods, relative to chance
The third demonstration turns a cue’s effect size into the detection accuracy it can support, showing why even a real cue leaves judges close to guessing.
From cue strength to detection accuracy
Suppose liars and truth-tellers differ on some behavioural cue by an effect size d. Even an ideal observer who used that cue perfectly would classify at only Φ(d/2). Slide dand watch how little accuracy a realistic cue buys.
Most real cues to deception have d near 0.2, which caps accuracy around 54% — almost exactly the average human performance found across hundreds of studies. Even a large cue (d = 0.5) reaches only about 60%, so no single behaviour can carry a confident judgement.
Neural Approaches
If deception has no autonomic signature, perhaps it has a neural one. Two techniques have pursued this idea. The first uses the P300, an event-related brain potential elicited when a person recognises a rare, meaningful stimulus. Applied to the Concealed Information Test logic, the P300 to a crime-relevant item should be larger in someone who recognises it, and early work proposed that this electroencephalographic signature could index guilty knowledge more directly than skin conductance (Farwell & Donchin, 1991). The approach inherits the strength of the recognition paradigm — it reads memory, not arousal — but also its precondition that investigators hold genuinely concealed details.
The second technique uses functional magnetic resonance imaging to look for brain regions more active during lying than truth-telling. Early studies reported that deceptive responses engaged prefrontal and anterior cingulate regions associated with cognitive control and response inhibition, consistent with the intuition that lying is effortful (Langleben et al., 2002). Subsequent work showed that the pattern depends heavily on the kind of lie — spontaneous versus rehearsed, memorised versus improvised — so that no single “deception centre” lights up across conditions (Ganis et al., 2003).
These findings are genuine, but the leap from group-averaged laboratory contrasts to a courtroom verdict about one person is enormous, and the field has been clear-eyed about it. A synthesis of behavioural, autonomic, and neural measures concludes that neuroimaging has not been shown to detect deception reliably in individuals, that its laboratory results do not transfer to realistic conditions, and that its methods are as vulnerable to countermeasures as the polygraph — a caution that has kept fMRI-based lie detection out of most courts (Meijer et al., 2016).
Worked Example
Consider the screening scenario the National Research Council warned about. Suppose an agency polygraphs 10,000 employees to find the rare spy, and suppose the base rate of spies is 1 in 100, so 100 of the 10,000 are guilty and 9,900 are loyal. Suppose too that the test is fairly accurate: it correctly flags 80% of spies (sensitivity 0.80) and correctly clears 80% of the innocent (specificity 0.80). How trustworthy is a “fail” result?
Of the 100 spies, the test flags 80% — 80 true positives — and misses 20. Of the 9,900 loyal employees, the test wrongly flags 20% — 0.20 × 9,900 = 1,980 false positives. So 80 + 1,980 = 2,060 people fail the test in total, but only 80 of them are actually spies. The positive predictive value is 80 / 2,060 = 0.0388, about 3.9%.
The result is counter-intuitive but exact: with these numbers, more than 96% of the people who fail the polygraph are innocent, and the agency would have to investigate 2,060 employees to catch 80 spies while still missing 20. This is not a flaw in the arithmetic but the unavoidable consequence of a low base rate — when the thing being screened for is rare, even a fairly accurate test generates false positives that swamp the true ones. It is precisely why the National Research Council judged the polygraph unsuitable for security screening, and the same logic constrains every lie-detection method proposed for a low-prevalence setting (National Research Council, 2003).
Discussion
The history of lie detection is a long collision between an appealing premise and an unyielding fact. The premise is that deception, being a departure from the truth, must leave a mark — on the body, the face, the voice, or the brain. The fact is that it does not leave a distinctive one: every mark it leaves is also left by other states, and the best a method can do is play the odds. Seen this way, the differences among methods are differences in how honestly they confront that fact. The polygraph’s Comparison Question Test is weakest because it pretends arousal is a proxy for deception when arousal has a dozen causes (National Research Council, 2003). The Concealed Information Test is strongest because it does not claim to detect lying at all — only recognition — and prices its errors explicitly (Ben-Shakhar & Elaad, 2003).
The behavioural tradition offers the field’s hardest lesson. After a century of looking, the cues are weak, judges are barely better than chance, and confidence and experience do not help (DePaulo et al., 2003; Bond & DePaulo, 2006). The productive turn has been to stop reading passive leakage and to make deception cognitively costly, so that the small differences that do exist are amplified rather than hunted for (Vrij et al., 2019). Running beneath all of this is the base-rate problem: no accuracy figure means anything without the prevalence of deception in the setting, and in the low-prevalence settings where lie detection is most tempting — security screening, mass interviewing — it is most misleading. The honest summary is that lie detection can shift the odds under favourable conditions but cannot deliver the certainty its name promises.
Current Directions
Two lines of contemporary work follow directly from these lessons. The first is the cognitive approach to interviewing, which reframes lie detection as an active task rather than passive observation. Because sustaining a lie under questioning consumes cognitive resources, interviewers can impose additional load — requesting fine-grained detail, asking for events in reverse chronological order, or springing unanticipated questions — and thereby enlarge the behavioural and verbal differences between liars and truth-tellers. Reviews of this programme report more promising accuracy than passive cue-reading, though the gains are still moderate and depend on skilled implementation (Vrij et al., 2010; Vrij et al., 2019).
The second line is a sustained methodological reckoning. The reappraisal of the cues-to-deception literature has prompted the field to treat older effect sizes as upper bounds inflated by small samples and selective publication, and to demand pre-registration and adversarial collaboration before a cue is believed (Luke, 2019). The same modesty has been applied to the physiological and neural methods, where a broad appraisal argues that behavioural, autonomic, and neural measures alike have been oversold, and that progress requires realistic base rates, countermeasure-resistant designs, and honest reporting of field performance rather than laboratory optima (Meijer et al., 2016). The trajectory of the field is thus away from the promise of a machine that reads lies and toward a disciplined accounting of how much, and under what conditions, the odds can be moved.
Common Misconceptions
- “The polygraph detects lies.”
- The polygraph records autonomic arousal — heart rate, respiration, skin conductance — not deception. Because arousal is produced by fear, anger, and stress as readily as by lying, the inference from the recording to a lie is exactly where the method is weakest (National Research Council, 2003).
- “Liars give themselves away with obvious body language.”
- The meta-analytic evidence shows behavioural cues to deception are weak and inconsistent; gaze aversion and fidgeting are not reliable signs. Untrained and trained observers alike judge lies at about 54% accuracy, barely above chance (DePaulo et al., 2003; Bond & DePaulo, 2006).
- “A high-accuracy test is trustworthy in any setting.”
- Accuracy is meaningless without the base rate. When deception is rare, even an 80%-accurate test flags far more innocent people than guilty ones, because the many innocents each carry a small false-positive risk that overwhelms the few true positives (National Research Council, 2003).
- “Brain imaging can now read lies directly.”
- Functional imaging reveals group-average differences between lying and truth-telling under laboratory conditions, but these depend on the type of lie, do not reliably classify individuals, and are vulnerable to countermeasures — which is why fMRI lie detection is not accepted in most courts (Ganis et al., 2003; Meijer et al., 2016).
Glossary
- Base rate.
- The prevalence of the condition being tested for in the population examined; when deception is rare, a low base rate makes even an accurate test generate mostly false positives.
- Comparison Question Test.
- The dominant field polygraph protocol, which compares a subject’s physiological reactions to relevant questions against reactions to broadly arousing comparison questions.
- Concealed Information Test.
- A physiological method that tests whether an examinee recognises crime-relevant details hidden among plausible alternatives, rather than whether they are lying; historically the Guilty Knowledge Test.
- Countermeasure.
- A deliberate physical or mental action — controlled breathing, mental arithmetic, a covert muscle contraction — used by an examinee to distort physiological responses and defeat a polygraph test.
- Cues to deception.
- Observable behavioural, verbal, or vocal signs proposed to distinguish liars from truth-tellers; meta-analysis finds them weak and inconsistent.
- Electrodermal response.
- A change in the electrical conductance of the skin driven by sweat-gland activity, the most diagnostic autonomic channel in the Concealed Information Test.
- Event-related potential.
- A brain electrical response time-locked to a stimulus, measured by electroencephalography; the P300 component indexes recognition of rare, meaningful items.
- Guilty knowledge.
- Information about a crime that only the perpetrator (and investigators) would possess; the basis on which the Concealed Information Test discriminates recognition from ignorance.
- Nonverbal leakage.
- The proposal that concealed emotions escape through poorly controlled channels such as the body and fleeting facial expressions, the historical premise of behavioural lie detection.
- Orienting response.
- The reflexive physiological reaction to a novel or significant stimulus; in the Concealed Information Test a guilty examinee orients more strongly to the recognised critical item.
- P300.
- A positive event-related potential occurring about 300 milliseconds after a rare, meaningful stimulus; proposed as a neural index of guilty knowledge.
- Polygraph.
- An instrument that records multiple autonomic channels — respiration, cardiovascular activity, and electrodermal response — while an examiner poses questions; colloquially the “lie detector.”
- Positive predictive value.
- The probability that a person who fails a test is actually deceptive; it falls sharply as the base rate of deception falls, even when test accuracy is high.
- Sensitivity.
- The proportion of genuinely deceptive examinees a test correctly flags; the complement is the miss rate.
- Specificity.
- The proportion of genuinely truthful examinees a test correctly clears; the complement is the false-positive rate that drives the base-rate problem.
Key Researchers
Gershon Ben-Shakhar
(1942–2026). Professor Emeritus of Psychology at the Hebrew University of Jerusalem and a leading figure in the Concealed Information Test tradition, whose meta-analytic work established the method’s validity and its strong protection of the innocent (Ben-Shakhar & Elaad, 2003). ORCID · Wikipedia · Wikidata
Paul Ekman
(1934–2025). Psychologist who pioneered the study of facial expression and, with Wallace Friesen, the theory of nonverbal leakage and microexpressions as clues to deception (Ekman & Friesen, 1969). Wikipedia · Wikidata
Charles R. Honts
(living). Professor Emeritus of Psychological Science at Boise State University, known for experimental work on the Comparison Question Test and on polygraph countermeasures (Honts et al., 1994). ORCID · Faculty · Google Scholar
William G. Iacono
(living). Regents Professor of Psychology at the University of Minnesota; a psychophysiologist and prominent scientific critic of the polygraph whose opinion surveys documented the field’s scepticism (Iacono & Lykken, 1997). ORCID · Faculty · Google Scholar · Wikipedia
Timothy J. Luke
(living). Associate Professor of Psychology at the University of Gothenburg, whose reappraisal of the cues-to-deception literature argued that such cues are far weaker than the canonical estimates suggest (Luke, 2019). ORCID · Faculty · Google Scholar
Bruno Verschuere
(living). Professor of Forensic Psychology at the University of Amsterdam, a Concealed Information Test and deception-detection researcher who has appraised the behavioural, autonomic, and neural methods together (Meijer et al., 2016). ORCID · Faculty · Google Scholar
Aldert Vrij
(living). Professor of Applied Social Psychology at the University of Portsmouth and the most cited empirical researcher on verbal and nonverbal lie detection; originator of the cognitive approach to interviewing (Vrij et al., 2010; Vrij et al., 2019). ORCID · Faculty · Wikipedia · Wikidata
Frequently Asked Questions
What is lie detection?
Lie detection is the attempt to distinguish truthful from deceptive statements by measuring physiological, behavioural, verbal, or neural signals rather than by gathering independent evidence. Its methods range from the polygraph and the Concealed Information Test to the analysis of speech and behaviour and, more recently, brain imaging. All share the same difficulty: there is no signal that occurs when and only when a person lies.
Does the polygraph actually detect lies?
No. The polygraph records autonomic arousal — heart rate, blood pressure, respiration, and skin conductance — not deception itself. Because that arousal can be produced by fear, anger, or the stress of being suspected, interpreting it as a lie is an inference that the U.S. National Research Council judged scientifically unsupported and unsuitable for high-stakes screening.
How accurate are people at detecting lies?
On average, about 54% — only slightly better than the 50% expected from guessing. A meta-analysis of hundreds of studies found this figure holds across untrained observers and experienced professionals alike, and that neither confidence nor experience reliably improves it.
What is the Concealed Information Test?
It is a physiological method that tests whether an examinee recognises details of a crime that only the perpetrator would know, presenting the true detail among plausible alternatives and measuring the orienting response to each. Because an innocent person reacts at chance, the test protects the innocent well and has a calculable, controllable false-positive rate.
Why does the base rate matter so much?
Because when the condition being screened for is rare, the many innocent people each carry a small chance of a false positive, and those false positives quickly outnumber the true positives. Even an 80%-accurate test used where only 1% are guilty will flag mostly innocent people, which is why accuracy figures are meaningless without the base rate.
Can polygraph tests be beaten?
Yes. Because the Comparison Question Test depends on a person reacting more to comparison than to relevant questions, an examinee can defeat it by covertly amplifying their reactions to the comparison questions through mental arithmetic, controlled breathing, or a discreet muscle contraction. Laboratory studies show such countermeasures substantially reduce accuracy and often escape detection.
Can brain imaging detect lies?
Functional MRI reveals group-average differences between lying and truth-telling in prefrontal and cingulate regions under laboratory conditions, but these patterns depend on the kind of lie, do not reliably classify individuals, and are vulnerable to countermeasures. For these reasons neuroimaging-based lie detection is not accepted as reliable in most courts.
What is the most promising modern approach?
The cognitive approach to interviewing, which makes lying mentally harder rather than searching for passive behavioural leaks. By imposing cognitive load — requesting detail, asking for events in reverse order, posing unanticipated questions — interviewers can enlarge the small differences between liars and truth-tellers, achieving better accuracy than reading body language, though the gains remain moderate.
References
Ben-Shakhar, G., & Elaad, E. (2003). The validity of psychophysiological detection of information with the Guilty Knowledge Test: A meta-analytic review. Journal of Applied Psychology, 88(1), 131–151. https://doi.org/10.1037/0021-9010.88.1.131
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
Ekman, P., & Friesen, W. V. (1969). Nonverbal leakage and clues to deception. Psychiatry, 32(1), 88–106. https://doi.org/10.1080/00332747.1969.11023575
Farwell, L. A., & Donchin, E. (1991). The truth will out: Interrogative polygraphy (“lie detection”) with event-related brain potentials. Psychophysiology, 28(5), 531–547. https://doi.org/10.1111/j.1469-8986.1991.tb01990.x
Ganis, G., Kosslyn, S. M., Stose, S., Thompson, W. L., & Yurgelun-Todd, D. A. (2003). Neural correlates of different types of deception: An fMRI investigation. Cerebral Cortex, 13(8), 830–836. https://doi.org/10.1093/cercor/13.8.830
Honts, C. R., Raskin, D. C., & Kircher, J. C. (1994). Mental and physical countermeasures reduce the accuracy of polygraph tests. Journal of Applied Psychology, 79(2), 252–259. https://doi.org/10.1037/0021-9010.79.2.252
Iacono, W. G., & Lykken, D. T. (1997). The validity of the lie detector: Two surveys of scientific opinion. Journal of Applied Psychology, 82(3), 426–433. https://doi.org/10.1037/0021-9010.82.3.426
Langleben, D. D., Schroeder, L., Maldjian, J. A., Gur, R. C., McDonald, S., Ragland, J. D., O’Brien, C. P., & Childress, A. R. (2002). Brain activity during simulated deception: An event-related functional magnetic resonance study. NeuroImage, 15(3), 727–732. https://doi.org/10.1006/nimg.2001.1003
Luke, T. J. (2019). Lessons from Pinocchio: Cues to deception may be highly exaggerated. Perspectives on Psychological Science, 14(4), 646–671. https://doi.org/10.1177/1745691619838258
Lykken, D. T. (1959). The GSR in the detection of guilt. Journal of Applied Psychology, 43(6), 385–388. https://doi.org/10.1037/h0046060
Meijer, E. H., Verschuere, B., Gamer, M., Merckelbach, H., & Ben-Shakhar, G. (2016). Deception detection with behavioral, autonomic, and neural measures: Conceptual and methodological considerations that warrant modesty. Psychophysiology, 53(5), 593–604. https://doi.org/10.1111/psyp.12609
National Research Council. (2003). The polygraph and lie detection. National Academies Press. https://doi.org/10.17226/10420
Vrij, A., Granhag, P. A., & Porter, S. (2010). Pitfalls and opportunities in nonverbal and verbal lie detection. Psychological Science in the Public Interest, 11(3), 89–121. https://doi.org/10.1177/1529100610390861
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