Deception

Abstract

Deception is a form of social behavior: the deliberate attempt by one party to create in another a belief the first party holds to be false, without forewarning. It is pervasive yet remarkably hard to catch, and the science of deception is organised around the gap between how confidently people believe they can spot a lie and how poorly they actually do. This article develops the main threads: the leakage hypothesis that deceit escapes through nonverbal behavior; the meta-analytic finding that behavioral cues are weak and few; the near-chance ceiling on lie-detection accuracy; the cognitive approach that improves detection by making lying more effortful rather than by reading better cues; Truth-Default Theory and the uneven distribution of lying across a few prolific liars; and the linguistic and indirect routes. A worked example and three interactive demonstrations make each manipulable.

Keywords: deception, social behavior, lie detection, nonverbal leakage, truth-default

Lying is one of the most ordinary acts of social life and one of the most consequential when it fails to be caught. People deceive to spare feelings, to gain advantage, to avoid punishment, and to manipulate; institutions from courtrooms to border crossings stake serious decisions on whether a person is telling the truth. And almost everyone believes they can tell—that a liar will fidget, avert their gaze, or betray themselves in some visible tell. The science of deception exists largely to test that belief, and its central finding is uncomfortable: the behavioral signs people trust are mostly myths, the real cues are faint, and human observers perform barely above chance.

What cognitive and behavioural science brings to the problem is a shift from asking what a liar looks like to asking why lying is so hard to detect and what, if anything, would help. The answer turns out to involve the weakness of the cues themselves, the biases of the observer, the cognitive demands lying places on the liar, and the default assumption of honesty that governs ordinary communication. The sections below take these in turn—leakage, the weakness of the cues, the accuracy ceiling, the cognitive approach, the truth-default, and the linguistic and indirect routes—after first placing deception within the broader category of social behavior under which it is classified.

What Deception Is

Deception, in the sense the behavioural sciences study it, is the deliberate attempt by one person to create in another a belief that the first person holds to be false. Three elements do the defining work. It is intentional—an honest mistake is not a lie. It involves a known falsehood—the deceiver believes the instilled belief to be untrue, whether or not it actually is. And it is done without forewarning—an actor or a magician, whose audience consents to being misled, does not deceive in the relevant sense. On this account deception is not a feeling or a trait but a goal-directed act, defined by the relation between the deceiver and the target rather than by any visible feature of the deceiver alone.

Deception is strikingly common. In a set of diary and survey studies of self-reported lying, people reported telling on the order of one to two lies a day on average—though, as the later sections show, that average is misleading because lying is concentrated in a minority of prolific liars (Serota et al., 2010). What makes deception a scientific puzzle rather than a simple matter of observation is that, for all its frequency, it is extraordinarily hard to detect: the behaviors people expect to betray a liar are largely absent, and those that do differ between liars and truth-tellers do so only faintly (DePaulo et al., 2003).

In the Medical Subject Headings vocabulary, Deception is a descriptor defined as the act of deceiving or the fact of being deceived, filed under Social Behavior. The placement is apt: deception is social through and through, defined entirely by the attempt of one party to shape the beliefs of another. The remainder of this article takes the mechanisms of deception and its detection in turn—leakage, the weakness of the cues, the accuracy ceiling, the cognitive approach, the truth-default, and the linguistic and indirect routes—and connects each to the cognition it requires.

Types of Deception

In the Medical Subject Headings vocabulary, Deception (D008192) is filed as a narrower descriptor of social behavior (D012919), and MeSH records one narrower descriptor beneath it. MeSH is an indexing classification built for retrieving the biomedical literature, not a theory of how deception is organised; its hierarchy reflects historical indexing practice rather than the psychological distinctions the experimental literature draws, and the two ways of carving the domain are orthogonal and should not be conflated.

This descriptor does not yet have its own article on this site, so the term is listed here but not linked. Its placement under Deception encodes a substantive point: disinformation is deception scaled up to the level of publics and institutions, sharing the defining feature of the parent construct—the deliberate instilling of a belief known to be false—while adding the dimensions of organised production and mass distribution. The study of individual deception and the study of disinformation increasingly inform one another, since the linguistic and psychological signatures that mark an individual lie also appear in fabricated information at scale (Markowitz & Hancock, 2014).

Perspectives on Deception

The science of deception is organised less by one grand theory than by a succession of perspectives, each tied to a distinct question. The earliest asked what betrays a liar and answered in terms of nonverbal leakage. A second, meta-analytic perspective asked how strong those cues actually are and found them faint. A third asked how well people detect lies and established a near-chance accuracy ceiling. A fourth, the most important recent synthesis on the applied side, asked what would improve detection and answered with a cognitive approach that makes lying harder. A fifth reframed the whole problem around the truth-default—the observation that people believe one another by default and that most lies come from a few prolific liars. These perspectives are layers rather than rivals: the same lie may leak faintly, resist detection, and be caught, if at all, only when the situation is engineered to expose it. Table 1 summarises them, and the sections that follow develop each in turn.

Table 1

Perspectives on Deception

Leakage and the search for behavioral cues

The modern science of deception began with an intuition made precise. Paul Ekman and Wallace Friesen proposed that when a person lies, the effort of concealment produces nonverbal leakage: because people control their faces more than their bodies and their words more than their faces, deceit suppressed in one channel escapes through another, leaving “clues to deception” in the hands, feet, and fleeting facial movements the liar does not think to manage (Ekman & Friesen, 1969). The leakage hypothesis gave the field its founding research programme: if deceit escapes through behavior, then cataloguing which behaviors betray it should make lies detectable.

Miron Zuckerman, Bella DePaulo, and Robert Rosenthal organised the early findings into a four-factor theory: lying is accompanied by heightened arousal, by the emotions of guilt and fear, by the greater cognitive effort of fabricating, and by liars’ attempts to control their behavior—and each of these should, in principle, surface as a detectable cue (Zuckerman et al., 1981). The framework was influential and productive, but it also set up the field’s central disappointment: when the predicted cues were measured across many studies, they proved far weaker than the theory’s intuitive appeal suggested. Demo 1 lets the reader set how often a candidate cue appears in liars versus truth-tellers and read off how little a faint difference reveals.

How weak the cues really are

The definitive test came from a meta-analysis by Bella DePaulo and colleagues that synthesised 158 cues to deception across more than a hundred studies. Its conclusion was sobering: most behaviors bore no reliable relation to lying at all, and the cues that did—liars sounding slightly more tense, telling slightly less compelling and less detailed stories, repeating themselves a little more—had small effect sizes and large overlap between liars and truth-tellers (DePaulo et al., 2003). Crucially, the cues people most trust—gaze aversion, fidgeting, nervousness—were among the least diagnostic. There is no Pinocchio’s nose: no single behavior, and no small set of them, reliably marks a lie.

Timothy Luke has pressed the point further, arguing that even the modest effect sizes in the cue literature are likely over-estimates, inflated by small samples, selective reporting, and analytic flexibility, so that the true diagnostic value of behavioral cues is probably smaller still (Luke, 2019). The implication reframes the whole enterprise: the problem is not that observers look at the wrong cues but that the cues themselves carry almost no information. If deception is to be detected, it will not be by reading a liar’s demeanour, because the demeanour of liars and truth-tellers is very nearly the same.

The accuracy ceiling

If the cues are weak, human performance should be poor—and it is. Charles Bond and Bella DePaulo meta-analysed more than 200 studies in which people judged whether others were lying or telling the truth, and found a mean accuracy of about 54%, where 50% is chance (Bond & DePaulo, 2006). The figure is strikingly stable: it hardly varies across observers, so that professionals such as police officers do little better than students, and confidence is largely unrelated to accuracy. Two further regularities stand out. People show a truth bias—they judge messages truthful more often than deceptive, so they are better at correctly classifying truths than lies. And the small variance that exists lies more in how detectable a given sender is than in how skilled a given observer is: some people are transparent liars, but almost no one is an expert lie-catcher.

This near-chance ceiling, built on the weakness of the cues, is the empirical heart of the field. It explains why confident folk beliefs about spotting liars persist despite being wrong—feedback in everyday life is rare and ambiguous—and it sets the benchmark any proposed improvement must beat. Demo 2 represents the truth-teller and liar as two overlapping distributions on a single behavioral axis and shows how, when the overlap is as large as the cue research implies, even an optimal observer is pinned near 54%.

The cognitive approach to lie detection

If liars and truth-tellers behave almost identically under ordinary questioning, one response is to change the questioning so that they no longer do. This is the premise of the cognitive approach to lie detection developed by Aldert Vrij and colleagues: lying is usually more cognitively demanding than telling the truth—the liar must invent a plausible account, keep it consistent, monitor the listener, and suppress the truth—so interventions that increase cognitive load, or that exploit the liar’s limited spare capacity, should degrade the liar’s performance more than the truth-teller’s and widen the difference between them (Vrij, Granhag & Porter, 2010). Techniques include asking interviewees to tell their story in reverse order, maintaining eye contact, or answering unanticipated questions—each harder for someone fabricating than for someone remembering.

A meta-analysis of the approach found that it raised classification accuracy appreciably above the standard-interview baseline, with correct-classification rates in the vicinity of 70% rather than the mid-50s (Vrij, Fisher & Blank, 2017). The gain is theoretically important because of where it comes from: not from teaching observers to read better cues, but from changing the task so that the liar produces more and stronger cues. Detection improves by acting on the sender, not the receiver. Demo 3 lets the reader impose cognitive load and watch the liar and truth-teller distributions pull apart, lifting accuracy off the 54% floor.

Truth-Default Theory and the prevalence of lying

Timothy Levine’s Truth-Default Theory reframes the whole problem from the receiver’s side. Its starting observation is that human communication runs on a presumption of honesty: people passively believe what they are told unless something actively triggers suspicion, and this truth-default is adaptive because most communication is in fact honest and the default makes efficient communication possible (Levine, 2014). The truth bias documented in the accuracy literature is, on this view, not a flaw but the sensible operation of the default. It also implies that lies are usually caught not in the moment, by reading demeanour, but later—through confessions, third-party information, physical evidence, and inconsistencies that surface over time.

The theory fits a striking fact about how lying is distributed. When Kim Serota, Timothy Levine, and Franklin Boster asked large samples how many lies they had told in the past 24 hours, the average was around one to two—but the distribution was heavily skewed: most people reported telling no lies at all on a given day, while a small number of prolific liars accounted for the bulk of the lies told (Serota et al., 2010). The “everybody lies all the time” cliché is, statistically, wrong: lying is the exception for most people and a habit for a few. That concentration matters for detection, because it means the base rate of deception in ordinary exchanges is low, which is exactly the condition under which a truth-default pays.

Linguistic and indirect routes

Two further lines of work look for deception where the demeanour cues fail. The first is linguistic: the words people choose, rather than how they move, may carry traces of deceit. David Markowitz and Jeffrey Hancock analysed the language of the fabricated scientific papers of the fraud Diederik Stapel and found measurable differences from his genuine work—the fraudulent papers used more terms conveying certainty and more science-related jargon, consistent with an effort to make invented findings sound convincing (Markowitz & Hancock, 2014). Automated text analysis, especially in computer-mediated communication where the message is all there is, has become an active route to detection that sidesteps the near-useless nonverbal channel.

The second line asks whether people register deception indirectly even when their explicit judgments hover at chance. Leanne ten Brinke, Kathleen Vohs, and Dana Carney reviewed evidence that indirect or unconscious measures—reaction-time and associative tasks—can discriminate lies from truths better than deliberate judgments do, suggesting the information is partly available to the perceiver but not well used by conscious deliberation (ten Brinke et al., 2016). The claim remains debated and the effects are modest, but it keeps open a genuine question: whether the accuracy ceiling is a limit on the information available or a limit on how deliberate judgment uses it.

Worked Example

The mechanisms become concrete in three small exercises, each matching one of the demonstrations. The indices below are illustrative—they capture the qualitative shape of published findings rather than reproducing any single reported statistic.

Cue diagnosticity (leakage). Suppose a candidate “tell”—say, a brief gaze aversion—appears in 30% of liars and 20% of truth-tellers. Its diagnostic value is the likelihood ratio, 0.30 ÷ 0.20 = 1.5: seeing the cue makes a lie only 1.5 times as likely as before, a weak signal. Starting from an even prior of 50/50, observing the cue updates the probability of a lie to only 0.30 ÷ (0.30 + 0.20) = 0.60. A cue most people would treat as damning moves the needle barely past a coin flip—and 0.30-versus-0.20 is already more separation than most real cues provide. Demo 1 lets the reader set the two rates and read off the resulting diagnosticity.

The accuracy ceiling (signal detection). Represent each sender by a value on one behavioral axis, with truth-tellers and liars drawn from two overlapping normal distributions separated by d′, the sensitivity of the behavioral signal. An observer who places the optimal threshold midway between them achieves accuracy Φ(d′ / 2), where Φ is the standard normal cumulative function. If the behavioral difference is as small as the cue research implies, d′ ≈ 0.2, then accuracy = Φ(0.1) = 0.540—54%, exactly the empirical ceiling. No change in the observer’s skill can raise this; it is fixed by the overlap of the two distributions.

The cognitive approach (shifting d′). Now suppose a cognitive-load interview makes lying markedly harder, pulling the two distributions apart to d′ ≈ 1.1. Accuracy becomes Φ(1.1 / 2) = Φ(0.55) = 0.709—about 71%, matching the gain the cognitive-approach meta-analysis reports. The lesson of the arithmetic is the lesson of the theory: the improvement comes entirely from increasing d′—making the liar behave more differently from the truth-teller—not from any better reading of a fixed signal. Demo 3 lets the reader slide d′ with the interview technique and watch accuracy climb off the floor.

Discussion

Read across its perspectives, the science of deception tells a coherent and somewhat deflationary story. The founding leakage hypothesis predicted that deceit would betray itself through behavior (Ekman & Friesen, 1969; Zuckerman et al., 1981), but the meta-analytic reckoning found the predicted cues faint, few, and easily confused with ordinary nervousness (DePaulo et al., 2003; Luke, 2019). Because the cues are weak, human accuracy is pinned near chance at about 54%, with a truth bias and little expertise to be found (Bond & DePaulo, 2006). The field’s most important constructive moves follow from taking this ceiling seriously rather than denying it.

One move acts on the sender: the cognitive approach raises detectability by making lying more effortful, lifting accuracy toward 70% not by better cue-reading but by producing stronger cues (Vrij, Granhag & Porter, 2010; Vrij, Fisher & Blank, 2017). The other reframes the receiver: Truth-Default Theory explains the truth bias as the sensible default of mostly-honest communication and relocates real detection to content, context, and evidence accumulated over time, while the prevalence data show that lying is concentrated in a few prolific liars rather than spread evenly across everyone (Levine, 2014; Serota et al., 2010). Linguistic analysis and indirect measures probe for the residue the demeanour cues miss (Markowitz & Hancock, 2014; ten Brinke et al., 2016). The honest summary is that catching a liar in the act, from how they look and sound, is close to impossible; what works is engineering the situation to make lying harder, and attending to what is said and what the evidence shows rather than to the myth of the telltale sign.

Current Directions

The most active applied front builds directly on the cognitive approach: rather than hunting for natural cues, researchers design interview protocols—asking for detail, posing unanticipated questions, requesting reverse-order recall—that actively elicit differences between liars and truth-tellers, and recent meta-analytic work has both consolidated the gains and mapped their boundary conditions (Vrij, Fisher & Blank, 2017). A second front follows deception out of the interview room and into text and networks, where automated linguistic analysis of reviews, messages, and fabricated content scales detection to settings no human observer could monitor, and where the boundary between individual deception and organised disinformation becomes a research question in its own right (Markowitz & Hancock, 2014). A third keeps alive the question of indirect detection—whether unconscious or implicit measures capture information that explicit judgments discard (ten Brinke et al., 2016). The common thread is a move away from the discredited search for a universal behavioral tell and toward methods—cognitive, linguistic, and contextual—that create or recover the information a bare lie does not reveal.

Common Misconceptions

Liars avoid eye contact and fidget.

These are the most widely believed cues and among the least diagnostic. The large cue meta-analysis found gaze aversion and fidgeting essentially unrelated to deception; the small reliable cues involve the content and quality of the account, not nervous mannerisms (DePaulo et al., 2003; Luke, 2019).

Trained professionals can spot lies reliably.

Across hundreds of studies, police, judges, and other professionals detect lies at roughly the same near-chance rate as students—about 54%. Experience and confidence are largely unrelated to accuracy (Bond & DePaulo, 2006).

Everybody lies constantly.

On average people report one to two lies a day, but the distribution is highly skewed: most people tell no lies on a given day, while a small group of prolific liars accounts for most of the lying. The mean is driven by the few, not the many (Serota et al., 2010).

The way to catch more liars is to watch them more closely.

Because liars and truth-tellers behave almost identically, closer watching does not help. Detection improves by changing the situation—imposing cognitive load so lying becomes harder—and by attending to content and evidence, not demeanour (Vrij, Granhag & Porter, 2010; Levine, 2014).

Glossary

Accuracy ceiling.

The robust finding that human observers distinguish lies from truths at about 54%, barely above the 50% expected by chance, with little variation across observers or training.

Cognitive approach.

A family of lie-detection techniques that improve accuracy by imposing cognitive load on the interviewee—reverse-order recall, unanticipated questions—so that lying, already effortful, becomes harder and more detectable.

Cognitive load.

The demand a task places on limited mental resources. Because lying is more effortful than telling the truth, deliberately raising cognitive load degrades a liar’s performance more than a truth-teller’s, which is the mechanism behind the cognitive approach.

Cue to deception.

A behavior hypothesised to differ between liars and truth-tellers. The meta-analytic finding is that such cues are few, weak, and mostly not the ones people believe in.

Deception.

The deliberate attempt by one party to create in another a belief the first party holds to be false, without forewarning; the subject of this article and a MeSH descriptor filed under social behavior.

Demeanour.

A person’s observable manner and behavior—expression, gaze, fidgeting, tone. The near-chance accuracy ceiling implies that reading demeanour is a poor route to detecting lies, which are better caught through content and evidence.

Disinformation.

False information deliberately created and spread to deceive, especially at scale; in MeSH a narrower descriptor beneath deception.

Leakage.

Ekman and Friesen’s proposal that deceit the liar tries to conceal escapes through less controllable channels—the body and fleeting facial movements—leaving clues to deception.

Likelihood ratio.

The probability of a cue given a lie divided by its probability given the truth; a measure of how much observing the cue should shift belief. Ratios near 1 indicate a near-useless cue.

Prevalence of lying.

The distribution of how much people lie. Self-report studies find a low average driven by a few prolific liars, with most people telling no lies on a given day.

Sensitivity (d′).

In signal-detection terms, the separation between the liar and truth-teller distributions on a decision axis. The weak cues to deception imply a very small d′, which fixes accuracy near chance.

Social behavior.

Action directed at or shaped by other members of one’s species; the MeSH parent category under which deception is filed, since deception is always defined relative to a target.

Truth bias.

The tendency to judge messages truthful more often than deceptive, which makes observers better at classifying truths than lies; in Truth-Default Theory, the sensible default of mostly-honest communication.

Truth-Default Theory.

Levine’s account that people believe one another by default, becoming suspicious only when triggered, and that lies are typically caught through content, context, and evidence rather than demeanour.

Key Researchers

Charles F. Bond

(Texas Christian University). Social psychologist whose meta-analysis with Bella DePaulo fixed the near-chance 54% ceiling on human lie-detection accuracy and documented its remarkable stability across observers. Google Scholar

Leanne ten Brinke

(University of British Columbia). Psychologist whose work on indirect and unconscious deception detection asks whether perceivers register lies through implicit measures better than through deliberate judgment. Faculty page

Bella M. DePaulo

(University of California, Santa Barbara). Social psychologist whose meta-analysis of behavioral cues to deception established that no cue is strongly or consistently tied to lying, overturning the intuitive leakage picture. ORCID · Google Scholar · Website

Paul Ekman

(1934–2025) (University of California, San Francisco). Pioneer of the scientific study of facial expression and deception whose leakage hypothesis, with Wallace Friesen, proposed that concealed emotion escapes through nonverbal channels and framed decades of cue research. Wikipedia · Google Scholar · Website

Pär Anders Granhag

(University of Gothenburg). Forensic psychologist whose work on strategic interviewing and the psychology of guilt and innocence helped establish the verbal and cognitive routes to lie detection. Faculty page · Google Scholar

Jeffrey T. Hancock

(Stanford University). Communication scientist whose research on linguistic cues to deception and lying in computer-mediated communication opened the automated, text-based route to detection. ORCID · Google Scholar · Wikipedia

Timothy R. Levine

(University of Alabama at Birmingham). Communication scientist whose Truth-Default Theory reframes deception around the presumption of honesty, and whose prevalence studies showed that lying is concentrated in a few prolific liars. ORCID · Google Scholar · Wikipedia

Aldert Vrij

(University of Portsmouth). Forensic psychologist who developed the cognitive approach to lie detection, showing that interviews which impose cognitive load raise accuracy by making lying harder rather than by reading better cues. Wikipedia · Faculty page

Frequently Asked Questions

What is deception in psychology?

Deception is the deliberate attempt by one person to create in another a belief the first person holds to be false, without forewarning. The three defining elements are intent, a known falsehood, and the absence of consent—which is why an honest mistake, or a magician’s illusion, is not deception in the relevant sense.

Can people really tell when someone is lying?

Not reliably from their behavior. Across hundreds of studies people detect lies at about 54% accuracy, barely above the 50% expected by chance, because the behavioral differences between liars and truth-tellers are very small. Confident beliefs about telltale signs are mostly unsupported.

Do liars avoid eye contact?

No. Gaze aversion is the most commonly believed cue and one of the least diagnostic. The large cue meta-analysis found it essentially unrelated to deception, along with fidgeting and other nervous mannerisms people trust.

What are the real cues to deception?

They are few and faint. The most reliable involve the content of the account—liars tend to give less detailed, less compelling, slightly more tense stories—rather than body language, and even these overlap heavily between liars and truth-tellers, so no cue is decisive.

What is the cognitive approach to lie detection?

It is a set of interview techniques—such as asking for the story in reverse order, posing unanticipated questions, or maintaining eye contact—that increase the cognitive load of the interview. Because lying is more effortful than truth-telling, load degrades the liar’s performance more and raises detection accuracy, often toward 70%.

What is Truth-Default Theory?

Developed by Timothy Levine, it holds that people believe one another by default and become suspicious only when something triggers doubt. This truth-default is adaptive because most communication is honest, and it implies that lies are usually caught through content, context, and evidence rather than by reading demeanour.

How often do people lie?

On average about one to two lies a day—but this average is misleading. The distribution is highly skewed: most people report telling no lies on a given day, while a small number of prolific liars account for most of the lies told.

Why are professionals no better at detecting lies?

Because accuracy is limited by the weakness of the cues, not by the observer’s skill or experience. Police, judges, and other professionals perform at about the same near-chance rate as students, and their greater confidence does not translate into greater accuracy.

References

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

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

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

Markowitz, D. M., & Hancock, J. T. (2014). Linguistic traces of a scientific fraud: The case of Diederik Stapel. PLoS ONE, 9(8), e105937. https://doi.org/10.1371/journal.pone.0105937

Serota, K. B., Levine, T. R., & Boster, F. J. (2010). The prevalence of lying in America: Three studies of self-reported lies. Human Communication Research, 36(1), 2–25. https://doi.org/10.1111/j.1468-2958.2009.01366.x

ten Brinke, L., Vohs, K. D., & Carney, D. R. (2016). Can ordinary people detect deception after all? Trends in Cognitive Sciences, 20(8), 579–588. https://doi.org/10.1016/j.tics.2016.05.012

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., 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

Zuckerman, M., DePaulo, B. M., & Rosenthal, R. (1981). Verbal and nonverbal communication of deception. Advances in Experimental Social Psychology, 14, 1–59. https://doi.org/10.1016/S0065-2601(08)60369-X