The Machine That Promised the Truth

In the summer of 1921, in a police station in Berkeley, California, a young officer named John Augustus Larson strapped a suspect to a contraption of rubber tubes and pressure cuffs and asked him a series of questions. A needle scratched its answer onto a slowly turning drum of paper. As the officer watched the line jump and settle, he believed he was watching something extraordinary: the physical trace of a lie, made visible for the first time in human history.

He was not. He was watching a frightened man’s heart beat faster. But the distinction, obvious as it seems now, would take the better part of a century to establish, and even that has not been enough to kill the idea. The dream Larson helped build is one of the most durable myths in modern criminal justice: that the truth lives in the body, that it can be extracted mechanically, and that a sufficiently clever instrument can drag it out into the light.

The strange thing is that the premise is half right. The body does betray liars. It leaks emotion through the face and the voice and the smallest involuntary movements. What it does not do is announce a lie cleanly, in a single unambiguous signal, the way a smoke detector announces a fire. And that gap, between the body that leaks and the machine that claims to read the leak, is where a hundred years of pseudoscience took root.

The psychologist who invented Wonder Woman

The story begins not with Larson but with a Harvard psychologist named William Moulton Marston. Around 1915, Marston became convinced that lying produced a measurable spike in systolic blood pressure. The reasoning was intuitive: deception is stressful, stress activates the body, and the body’s activation shows up in the cardiovascular system. Marston wrapped a cuff around the arms of his test subjects, asked them questions, and recorded the numbers as they climbed. ¹

Marston was a genuine scientist, but he was also a showman and a man of expansive obsessions, and the arc of his career bends in a direction almost too perfect to be true. In the 1940s, under the pen name Charles Moulton, he created the comic-book character Wonder Woman. Her signature weapon was the golden Lasso of Truth, a rope that forced anyone bound by it to speak honestly. This was not a coincidence. It was the same dream Marston had chased in the laboratory, transmuted into myth: a tool that could compel honesty, a device that made lying physically impossible. ²

Marston never built a working lie detector himself, and his blood-pressure test was crude. But he supplied the seductive idea at the center of the whole enterprise: that honesty and deception are opposite physiological states, and that a needle on a chart could tell them apart.

John Larson turned the idea into an instrument. In 1921, working under the reform-minded Berkeley police chief August Vollmer, Larson combined continuous measurements of blood pressure, pulse, and respiration into a single machine and set it to record all three at once onto a moving strip of paper. He called his invention the cardio-pneumo-psychogram. The name did not survive, but the device did, and it eventually acquired the word we still use: polygraph, from the Greek for “many writings.” ³

Larson intended his machine as a scientific instrument, a way to make interrogation more humane and more objective than the beatings and coercion that passed for police work at the time. His assistant, Leonarde Keeler, would go on to refine the device and, crucially, to sell it, turning the polygraph into a commercial product and a courtroom spectacle. By the 1930s, the machine had a mystique that outran anything its inventors could defend. Larson himself grew disillusioned. Late in life he called the polygraph “a Frankenstein’s monster,” a creation that had escaped his control and taken on a life the science could not support. ⁴

What the needle actually measures

The theory behind the polygraph sounds airtight until you look at it closely. Lying, the argument goes, is cognitively and emotionally taxing. To sustain a falsehood, a person must suppress the truth, invent a plausible alternative, monitor the listener’s reaction, and manage the fear of being caught. All of this triggers the sympathetic nervous system: the ancient fight-or-flight machinery that floods the body with adrenaline. The heart pounds. The palms sweat. Breathing turns shallow and quick.

A polygraph records exactly these changes. It tracks heart rate and blood pressure, respiration, and, most importantly, the electrical conductance of the skin, which rises as the sweat glands activate. What it produces is a detailed physiological portrait of arousal.

And there is the problem, stated in a single word. The machine detects arousal. It does not detect lying. Nothing in the biology distinguishes the racing heart of a guilty person from the racing heart of an innocent one who is terrified of being wrongly accused. The alarm system that the polygraph eavesdrops on rings for fear, for anger, for indignation, for embarrassment, for the simple stress of being wired to a machine in a room full of strangers who suspect you of a crime. The signal is real. Its meaning is not.

This is the fatal flaw, and it cuts in two directions at once. A truthful person, frightened and cornered, can produce a physiological response that looks exactly like guilt, and fail. Meanwhile a practiced liar, or someone who simply feels little anxiety about deception, can stay calm enough to sail through. The people the machine is worst at catching are precisely the ones society most wants caught: the composed, the callous, the habitual deceiver who has stopped feeling the sting of the lie.

The scientist who spent his life debunking it

No one attacked the polygraph more persistently than David T. Lykken, a psychologist at the University of Minnesota who spent decades documenting its failures. Lykken argued that the standard polygraph procedure, known as the Control Question Test, rested on assumptions that could not be justified. The test compares a subject’s reaction to relevant questions (“Did you steal the money?”) against reactions to vaguely threatening control questions (“Have you ever taken something that wasn’t yours?”). The theory holds that a guilty person will react more strongly to the relevant questions and an innocent person more strongly to the controls. But there is no scientific reason this should reliably be true, and Lykken assembled the evidence to show it often is not. ⁵

Lykken’s larger point was that the polygraph could not distinguish a specific memory of guilt from generalized emotional response. He championed an alternative approach, the Guilty Knowledge Test, which tried to detect whether a suspect recognized details of a crime that only the perpetrator would know. That method has more defensible logic, but it is rarely used, because it requires crime details that have been kept secret and does not fit the accusatory style of most interrogation.

The estimates of polygraph accuracy vary wildly depending on who is measuring and how, which is itself a symptom of the problem. Field studies and laboratory studies disagree; proponents cite figures above 90 percent, critics find results closer to chance. When the evidence for a diagnostic tool ranges from “highly accurate” to “barely better than a coin flip” depending on the study, the honest conclusion is that the tool is not measuring what it claims to measure.

In 2003, the matter received its most authoritative review. The National Research Council, an arm of the U.S. National Academy of Sciences, convened a panel of experts to examine the entire scientific literature on polygraph testing. Their report was withering. The panel concluded that the theoretical basis of the polygraph was weak, that its accuracy in real-world conditions was overstated, and that the evidence for its validity was “scanty and scientifically weak.” ⁶

Most pointedly, the council warned against using polygraphs for national-security screening, the exact purpose to which the U.S. government was, and still is, devoting enormous resources. In a screening context, where the vast majority of people tested are innocent, even a small false-positive rate produces a flood of wrongly accused loyal employees, while a determined spy who has learned to beat the machine slips through. The math is unforgiving, and the council spelled it out.

The truth that does leak

Here the story takes a turn that surprises most people, because the collapse of the polygraph does not mean the body keeps its secrets. It means the machine was listening to the wrong channel. While the polygraph was failing, a very different line of research was finding real, if fragile, signals of deception in the human face and voice.

The central figure here is Paul Ekman, a psychologist who spent decades studying facial expression frame by frame, slowing film down until he could see what happens in the muscles of the face in fractions of a second. Ekman documented what he called microexpressions: involuntary flashes of genuine emotion that cross the face before a person can suppress them, sometimes lasting as little as a twenty-fifth of a second. A witness claiming grief might reveal a flicker of contempt; a suspect professing calm might betray a burst of fear. The mask slips, the truth escapes, and then the mask returns, often before the person is even aware of what happened. ⁷

Ekman’s work built on a deeper anatomical fact. Some of the muscles in the human face are not fully under voluntary control. The classic example is the genuine smile, which Ekman named the Duchenne smile after the 19th-century French physician who first described it. A real smile of pleasure engages the muscles around the eyes, the orbicularis oculi, producing the characteristic crinkle. A posed or social smile typically involves only the mouth. Most people cannot consciously activate the eye muscles on command, which is why a forced smile so often looks hollow: the eyes give it away. ⁸

Beyond the face, researchers catalogued other tells. Contrary to folklore, liars often move less rather than more, holding themselves unnaturally still as they concentrate on controlling their story and suppressing the fidgeting they imagine will give them away. The voice tends to climb in pitch under stress as tension tightens the vocal cords. Speech may slow or grow strangely careful. Verbal accounts of fabricated events tend to contain fewer sensory details and less spontaneous self-correction than accounts of things that actually happened.

So the body does betray liars, after a fashion. The truth leaks through the cracks of the performance. This is the grain of reality that has kept the whole enterprise alive for a hundred years, the reason the myth refuses to die.

The tell that isn’t there

And yet here is the finding that almost nobody wants to hear, the one that undoes the folklore even as it confirms it. No single tell reliably reveals a lie. Not one. There is no gesture, no expression, no vocal shift, no bead of sweat that means, dependably, across people and situations, that someone is lying.

The research on human lie detection is sobering. When ordinary people are asked to judge whether strangers are telling the truth, they perform at around 54 percent accuracy, barely above the 50 percent they would achieve by flipping a coin. A large meta-analysis by the psychologists Charles Bond and Bella DePaulo, pooling hundreds of studies, arrived at almost exactly this figure and found that it held stubbornly across professions. ⁹

More unsettling still, the people whose jobs depend on catching liars, police officers, customs agents, judges, are on average no better than the rest of us, and are frequently more confident while being no more accurate. Their confidence tends to attach to the very cues that fail: eye contact, fidgeting, nervousness, the gaze that shifts away. These are the cues everyone thinks reveal deception, and they are close to worthless, because they mark anxiety, not dishonesty, and anxiety is exactly what an innocent person under suspicion feels.

The microexpression, the Duchenne smile, the climbing pitch: each is a real phenomenon, but each is probabilistic, not diagnostic. It shifts the odds a little. It does not settle the question. Treat any one of them as proof and you will convict the innocent and exonerate the guilty with roughly equal enthusiasm.

What actually catches a liar

If no single signal works, what does? The answer, drawn from the research of investigators like Aldert Vrij, is disarmingly unglamorous. The most reliable route to detecting deception is not physiological at all. It is cognitive, and it is verbal. It lives in the structure of the story, not in the sweat on the brow. ¹⁰

The insight is that lying is hard work, and the difficulty can be exploited. A person recounting a real experience is reading it off memory. A person fabricating must construct and maintain a consistent alternative in real time while monitoring for contradictions. This is far more taxing, and the strain shows up not in blood pressure but in the coherence of the account. Skilled interviewers increase that cognitive load deliberately: asking for the story in reverse chronological order, requesting unexpected details, returning to the same events from different angles. Truthful accounts tend to survive this pressure, growing richer with detail. Fabricated ones tend to fracture, shedding specifics and contradicting themselves.

The real signal, in other words, is inconsistency. It is the story that shifts, the detail that crumbles when gently probed, the timeline that will not stay fixed. And detecting it requires no wires, no cuffs, no rolling drum of paper. It requires a patient, attentive human listener who is willing to hold two versions of an account side by side and notice where they fail to align.

This is why courts in most of the world refuse to admit polygraph evidence, and why the U.S. legal system treats it with deep suspicion. The machine does not measure what the law needs measured. It measures fear, and fear is not guilt.

The lie detector, it turns out, was never a machine. Marston’s Lasso of Truth was always fiction, and so, in a quieter way, was the polygraph that his idea inspired. The body genuinely does leak, but not in one clean signal a needle can trace. The truth hides in patterns: in the accumulation of small inconsistencies, in the shape of a story told and retold, in the gap between what a person claims and what their account can actually sustain. The next time someone swears they are being honest, the thing to watch is not the sweat on their brow. It is the architecture of what they are saying, and whether it holds together when you press on it.

Sources

1. Marston, W. M., “Systolic blood pressure symptoms of deception,” Journal of Experimental Psychology, 1917. — https://psycnet.apa.org/record/1926-02774-001
2. Lepore, Jill, The Secret History of Wonder Woman, Knopf, 2014. — https://www.penguinrandomhouse.com/books/226043/the-secret-history-of-wonder-woman-by-jill-lepore/
3. Alder, Ken, The Lie Detectors: The History of an American Obsession, Free Press, 2007. — https://www.simonandschuster.com/books/The-Lie-Detectors/Ken-Alder/9780803259881
4. American Psychological Association, “The Truth About Lie Detectors (aka Polygraph Tests),” 2004. — https://www.apa.org/topics/cognitive-neuroscience/polygraph
5. Lykken, David T., A Tremor in the Blood: Uses and Abuses of the Lie Detector, Plenum Press, 1998. — https://link.springer.com/book/10.1007/978-1-4899-6067-9
6. National Research Council, The Polygraph and Lie Detection, National Academies Press, 2003. — https://nap.nationalacademies.org/catalog/10420/the-polygraph-and-lie-detection
7. Ekman, Paul, Telling Lies: Clues to Deceit in the Marketplace, Politics, and Marriage, W. W. Norton, 1985. — https://wwnorton.com/books/9780393337457
8. Ekman, P. & Friesen, W. V., “Felt, false, and miserable smiles,” Journal of Nonverbal Behavior, 1982. — https://link.springer.com/article/10.1007/BF00987191
9. Bond, C. F. & DePaulo, B. M., “Accuracy of Deception Judgments,” Personality and Social Psychology Review, 2006. — https://journals.sagepub.com/doi/10.1207/s15327957pspr1003_2
10. Vrij, A., Fisher, R. & Blank, H., “A cognitive approach to lie detection,” Legal and Criminological Psychology, 2017. — https://bpspsychub.onlinelibrary.wiley.com/doi/10.1111/lcrp.12088

Related reading

• The Arithmetic of Cancer
• The Alarm That Feels Like Suffocation

More from the Body edition →