Can You Tell if I'm Lying?

by Brian Dunning
June 14, 2022

Maybe you're at the poker table, looking to see if your opponent is bluffing — a scrap of insight that could be worth thousands of dollars. Maybe you're a cop in the interview room watching the suspect for cues that will betray his guilt. Maybe you're a TSA agent in the airport watching people go by, looking for the specific little behaviors that reveal this person is up to something deceptive. Whichever it is, they're all part of the same craft, the observation of body language that will give away deceptive behavior. It's used in law, it's used in security, it's even used in sales; it's taught in trade schools and it's sold as learning products. So maybe it's time that we had a quick look to see if the science behind it is valid. In controlled testing, are these marvelous techniques any better than random chance at telling whether someone is lying?

A lot of people think so, and they've invested considerable money based on this notion. One of these is the United States' TSA (Transportation Security Administration). In 2007, the TSA deployed a program called SPOT, which stands for "Screening of Passengers by Observation Techniques" (it has since been renamed BDA, for "Behavior Detection and Analysis"). TSA has trained thousands of Behavior Detection Officers at a cost of some $900 million as of 2015. The system involves a 92-point checklist largely based on behaviors that are stereotypically associated with lying and deception: avoiding eye contact, fidgeting, touching their face, faster breathing, increased rate of blinking, etcetera — basically everything you've ever heard of as being a trait of liars.

SPOT's record has been abysmal. TSA does publish their statistics, and over one three-year period, their behavior detection pulled some quarter of a million people for secondary screening. Almost all were found to be not up to anything. Only about 3/4 of a percent were arrested, nearly all for completely unrelated charges like outstanding warrants — about the same rate you'd get from the general public. Not one has ever been linked to terrorism.

So in 2010, the Government Accountability Office determined that TSA had deployed SPOT without first validating its scientific basis. They spent the next two years analyzing the data from SPOT, and then in 2013, presented their report, titled TSA Should Limit Future Funding for Behavior Detection Activities. The Department of Homeland Security (TSA's parent agency) did not concur with the report, claiming that the GAO had not considered all the available research, and that the research it had considered lacked validity. It was a classic back-and-forth case of "My peer-reviewed journal article is better than yours." And so nothing happened — and the program continues today under its current name BDA.

So, whose peer-reviewed journal articles were better? To answer this, we can open up the topic to the broader picture. At the time the GAO made its determination, there was already a robust body of work on this topic. Two of the most often cited publications included the academic psychology textbook Detecting Lies and Deceit: Psychology of Lying and the Implication for Professional Practice by Aldert Vrij which cited more than 1000 pieces of research. Vrij found that there is not a single reliable non-verbal cue to lies or deceit.

The other was an important meta-analysis in the journal Personality and Social Psychology Review titled "Accuracy of Deception Judgments" by Charles Bond and Bella DePaulo. They looked at the results of 206 studies involving over 24,000 people and found that, with no special training, we can't tell lying deceptive people from honest non-deceptive people any better than random chance. Overall, people correctly identified lies only 47% of the time, and correctly identified the truth 61% of the time. Moreover, the data showed that supposedly trained professionals, like police investigators, psychiatrists, job recruiters, and presumably TSA Behavior Detection Officers, score no better than untrained laypeople.

One of these authors, Bond, also founded the Global Deception Research Team consisting of 91 academic psychologists located in 58 different countries. Each of them then performed a survey of an equal number of male and female subjects, asking them all (in the native language) what are the cues that let you know someone is lying. 2,320 respondents gave 11,157 responses, which coders then boiled down to 103 distinct beliefs. The leading answer by far, common to two thirds of respondents, was averting one's gaze. All other answers were less than half as common, but in order, they were:

• Nervousness
• Incoherence
• Body movements
• Facial expression
• Inconsistencies
• Uhs
• Facial color
• Pauses
• Arm, hand, and finger movements
• Changes in speech rate
• Noises like sighs
• Necessary to know the person
• Tone of voice
• Eye movements
• Sweating
• Fidgeting with hair, clothes, objects
• Unspecified changes in behavior
• Weak arguments and logic

So it can be fairly argued that these are the most commonly believed tells, or cues, that someone is lying or being deceptive. Most appear on the TSA SPOT checklist. And according to those more than 200 studies putting each of these to the test, not a single one is correct more often than random chance. You would literally fare just as well flipping a coin to tell whether a person is a terrorist in an airport, or a lying suspect in the interrogation room.

But if you watch old TV shows, you might have a different idea. There was a TV show from 2009-2011 called Lie to Me about an expert played by Tim Roth who would go around solving mysteries for law enforcement and other clients all by reading body language and subtle, split-second facial expressions called microexpressions. It was based on the work of controversial psychology researcher Paul Ekman who, in the 1970s, co-created a system he called the "facial action coding system", which considered microexpressions as indicators of deceit, among other things. Ekman's work was also a major underpinning for TSA's SPOT program, and accordingly, has faced much of the same criticism: a lack of experimental replication. In the face of all the rejection from his peers in psychology, Ekman stopped submitting his work for publication in academic journals years ago, bizarrely claiming national security concerns.

Remember those stats from the SPOT program, a quarter of a million detained, less than 1% arrested, all for unrelated stuff and none for terrorism? Here is how Ekman himself framed that in an article for the Washington Post:

Preliminary findings show that the overwhelming number of those who are taken out of line and detained for further investigation were intending to commit or had committed some kind of wrongdoing: They were wanted criminals, drug smugglers, money smugglers, illegal immigrants — and, yes, a few were suspected terrorists.

Unsurprisingly, Ekman's website — the Paul Ekman Group — sells all kinds of training programs and courses, and there's even a big 15% coupon at the top of the page.

Ekman's done a lot of good and important work in the field of psychology, but it's his vast body of work in detecting deception that has made him stand out. For as much academic criticism as it's received, it's been warmly embraced by pop-culture. The TV series Lie to Me and TSA's SPOT were hardly the only programs. The animation studio Pixar engaged him as a consultant on the movie Inside Out. The BBC made a series hosted by John Cleese called The Human Face which promoted his claims. And he has written some 15 mass market books with major publishers, guaranteeing that his fringe perspective on microexpressions as indicators of deception will continue to be the dominant one for decades to come.

Ekman and other supporters of behavior-based lie detection have raised some interesting defenses. One thing they point out is that nearly all of this research, which is conducted mostly at universities, relies on undergraduate students as the test subjects. Trained professionals like police detectives, on the other hand, are far more experienced at spotting deception. They also contend that in such experiments, students are told to lie about something dumb like what color eyes they have, things that are unimportant; while in the real world, liars might be concealing a murder or some other crime for which there could be grave consequences, and so they react with more obvious cues.

However, these defenses are really only persuasive to laypeople and to the mass media. Psychology researchers have always been well aware of issues such as these and have controlled for them. For example, one 2019 study published in the Journal of Investigative Psychology and Offender Profiling trained people using Ekman's own METT (micro-expressions training tool), which is, yes, for sale on his website. They watched videos of people either lying or telling the truth, in both real high-stakes and low-stakes environments. All groups performed slightly worse than random chance, both before and after the training with METT. In contrast with the claims made on Ekman's website, the authors concluded:

Our findings do not support the use of METT as a lie detection tool. The METT did not improve accuracy any more than a bogus training protocol or even no training at all. The METT also did not improve accuracy beyond the level associated with guessing.

In 2020, an opinion article was published in Frontiers in Psychology by a pair of psychology researchers who had been swimming in these waters for years and were just about fed up with how thoroughly the idea of tells had been debunked in recent decades, and yet it's still used and believed in worldwide. They noted that in 2016 alone, at least 206 scientific papers were published making the same old finding. They concluded:

Is the rational course simply to drop this line of research? ...We have now sufficient evidence that there are no specific non-verbal behavioral signals that accompany lying or deceitful behavior. We can safely recommend that courts disregard such behavioral signals when appraising the credibility of victims, witnesses, and suspected offenders. For psychology and law researchers it may be time to move on.

And so, that is perhaps where we will leave the topic too. It is time to move on. The detection of deception via behavioral cues, be they body language or microexpressions, may have a place somewhere; but if it does, researchers have not convincingly found it yet.

For our episode on lie detection via polygraphy, go here.

By Brian Dunning
Follow @BrianDunning