You catch your colleague glancing at their phone during a meeting. Their laugh at the joke that follows sounds a little too casual. The explanation they give for being late feels slightly off. And just like that, your brain has already built a story about what they're hiding.

But here's the uncomfortable truth: you almost certainly got it wrong.

The scientific literature on lie detection is quietly devastating. Study after study shows that the instincts we rely on when we suspect someone of dishonesty are, at best, barely better than flipping a coin. And at worst, actively misleading us.

The average person can detect lies with only around 54% accuracy [2]. That is not a typo. The average person, over the course of a lifetime of navigating relationships, jobs, and everyday transactions, has honed a set of lie-detection skills that amount to chance. Fifty-four percent is what you get when you have no idea what you're doing and you're guessing.

Even trained professionals perform only marginally better. Law enforcement officers, border security agents, and Secret Service personnel who have spent years studying deception achieve only around 64% accuracy in lie detection, even after formal training [2]. A 2007 peer-reviewed review of 50 years of research found no scientific evidence supporting voice analysis as a reliable indicator of deception [1].

Why are we so bad at this?

The answer lies in how human social cognition evolved. Most people operate by default on the assumption that others are telling the truth [1]. This is not naivety. It is strategy. Trust is the bedrock of social life at every level, from marriage to international commerce. A society where everyone assumed deception was the norm would collapse under the weight of its own suspicion.

So when your colleague laughs too hard at that joke, your brain does something remarkable: it overrides the suspicious signal and decides, by default, that they are simply being polite or awkward, not lying. This reflex is so deeply embedded that most of us never notice it.

Which brings us to one of the most pervasive myths in law enforcement and popular psychology.

The Body Language Myth

Walk into any police station or security training facility from the last forty years and you will find walls covered with charts on posture, eye contact, and fidgeting. The assumption is that liars betray themselves through non-verbal channels: they avoid eye contact, they touch their faces, they shift in their seats.

Here is what the research actually says: there is no evidence to substantiate that non-verbal lie detection, such as by looking at body language, is an effective way to detect lies [1]. This is not a controversial finding. It is the consensus of decades of experimental work.

The U.S. Supreme Court put it bluntly in United States v. Scheffer in 1998. Regarding polygraph evidence, the Court noted it was little better than could be obtained by the toss of a coin [3]. The American Psychological Association has stated that most psychologists agree there is little evidence that polygraph tests can accurately detect lies [3].

The gap between what pop psychology tells us and what science actually supports is staggering. An innocent person in an interrogation room may appear nervous, avoidant, and physically agitated exactly because they are terrified of not being believed. A practiced liar may project complete calm. Signs of emotion are not necessarily signs of guilt [2]. And non-verbal cues such as avoiding eye contact and fidgeting are not consistent indicators of lying, despite what countless television detectives would have us believe [1].

What the Polygraph Gets Wrong

The polygraph has been a fixture of American law enforcement for over a century. The device measures and records several physiological indicators, including blood pressure, pulse, respiration, and skin conductivity, while the subject is asked questions [3]. The theory is straightforward: lying raises arousal, and that arousal leaves a measurable signature.

The science disagrees. The American Psychological Association states that most psychologists agree there is little evidence that polygraph tests can accurately detect lies [3]. A comprehensive 2003 review by the National Academy of Sciences concluded there was little basis for the expectation that a polygraph test could have extremely high accuracy [1]. The U.S. Supreme Court has noted that polygraph evidence is little better than could be obtained by the toss of a coin [3].

The problem is that arousal is not deception. A truthful person with severe anxiety will show all the physiological markers of a liar. A well-prepared deceiver who has practiced breathing control techniques or mental arithmetic can defeat a polygraph without lying at all [3]. Even trained polygraph operators cannot reliably distinguish between deceptive and truthful responses [3].

This leaves us with a paradox. The polygraph persists in law enforcement, popular culture, and the public imagination because it feels scientific. It produces a number, a graph, something that looks like evidence. But the numbers do not measure what they claim to measure.

Microexpressions: The Half-Second Window

If body language fails us and the polygraph is pseudoscience, is there anything that actually works?

One of the most promising research areas involves microexpressions, brief involuntary facial movements that occur when the amygdala responds to stimuli before conscious intervention can suppress them [2]. Paul Ekman, the pioneering psychologist who developed the Facial Action Coding System, showed that these expressions flash across the face in less than half a second and express seven universal emotions: disgust, anger, fear, sadness, happiness, contempt, and surprise [4].

Microexpressions cannot be controlled, which makes them potentially valuable for detecting concealed emotions [2]. A person trying to hide their anger will briefly show it before their face rearranges into something more socially appropriate. This moment of leakage, though brief, may be the closest thing to a genuine tell that science has identified.

Ekman's work inspired the television series "Lie to Me" and the development of training tools like the MicroExpression Training Tool, which can improve detection accuracy from around 54% to about 64% [2][4]. But even trained professionals are wrong more than a third of the time. Microexpressions show emotion, not necessarily deception. A genuinely surprised person and a deceptive one can show the same microexpression. The window is narrow, and the interpretation is far from straightforward.

Words That Lie Does Not Use

If you want to catch a liar, try listening more carefully than you watch.

Research on verbal cues to deception has identified several patterns worth knowing about. Liars tend to give briefer and more vague statements that are detached from the self [1]. They use fewer first-person pronouns like I, me, and my. They speak in shorter statements with more pauses, and they may be slower to respond because lying requires more cognitive load [1].

The content of what they say is also revealing. Liars tend to avoid amending their statements or correcting themselves naturally. Truthful people will add clarifying details and self-correct in ways that liars, sticking to their rehearsed story, do not [1].

Highly structured stories with fewer details can also signal deception. A liar's account lacks the contextual and sensory detail that a genuine memory naturally contains. They lower the potential risk of inconsistency by saying less, but that very emptiness becomes its own signal [1].

None of these patterns are definitive. But in aggregate, and combined with what you already know to be true, they can help shift the odds slightly in your favour.

A Sceptical Frame of Mind

The evidence is clear: most people are poor lie detectors, and the tools we have been taught to rely on are not reliable. The good news is that the research also points toward what does work, even if the effects are modest.

Start with your own default setting. Most people assume others are truthful. This is normal. It is also your biggest vulnerability. Cultivating a slight sceptical frame of mind, without tipping over into paranoid suspicion, is perhaps the most useful starting point.

Pay attention to what people say and whether it holds together over time. Liars avoid amending their stories. They give less detail. They detach themselves from their own accounts linguistically. Ask follow-up questions that require elaboration. A truthful person will naturally add contextual detail. A liar will tend to deflect or give vague non-answers.

Watch for emotional leakage, particularly microexpressions that do not match the situation. But do not over-interpret. A nervous person is not necessarily lying. A calm person is not necessarily telling the truth.

And when it comes to formal lie detection tools, treat them with deep scepticism. The polygraph has been rejected by courts for good reason. Body language analysis has no scientific basis. Even microexpression analysis, for all its genuine scientific pedigree, is not a lie detector. It is, at best, an emotion detector, and emotions do not always mean deception.

The best lie detector we have is still a curious, sceptical, and socially intelligent mind. Train that instead.