The Junk Science Cops Use to Decide You’re Lying

The training session was billed as “cutting edge,” and dozens of law enforcement professionals signed up to learn about “New Tools for Detecting Deception” from a human lie detector who calls herself “Eyes for Lies.” Her real name is Renee Ellory, and she claims that she’s one of just 50 people identified by scientists as having the ability to spot deception “with exceptional accuracy.”

A flyer for the event, hosted by Wisconsin’s High Intensity Drug Trafficking Area — a federal program that supports law enforcement drug interdiction work — was included among a trove of law enforcement documents that were hacked and posted online in June under the title BlueLeaks. The promo copy leans heavily into Ellory’s skill at ferreting out deception in others. She is “exceptional at pinpointing a liar and can tell you why she doesn’t trust someone on the spot,” it reads. Training participants would learn how to “identify anger, contempt, and disgust before words are even spoken.” Course objectives were broad: Learn to differentiate between “real” and “fake emotional displays”; “recognize hidden emotions”; identify the “ways our subconscious brain leaks information when we lie”; “analyze body language that indicates deception”; gain tips to use when interviewing a psychopath; “identify the key features of expressions that reveal danger for you!”

Participants spanned the law enforcement spectrum and included the chief of a small police department, corrections officers, university cops, state troopers, various members of the Milwaukee Police Department as well as individuals from the U.S. Probation Office and the FBI. In surveys filled out after the training, which took place in November 2015, the common complaint was that there weren’t enough structured breaks; as one participant put it, “the mind can only absorb what the buttocks can tolerate.” But otherwise, a majority of the 82 respondents gave the training high marks. Participants wrote that they would incorporate what they’d learned into their police work. A number of them said the most valuable thing they learned was “the seven universal facial expressions that all people have all over the world as a good indicator” of lying, as one trainee put it.

It might seem reassuring that so many law enforcement officers found a skills training so valuable. But not in this case. That’s because Ellory’s lie detection training is based what many psychologists say are largely discredited theories, if not simply junk science. “It’s completely bogus,” said Jeff Kukucka, an assistant professor of psychology and law at Towson University who studies forensic confirmation bias, interrogations, and false confessions. “And what’s maybe more alarming about it … is that this isn’t new. We’ve known for quite a while that this stuff doesn’t work, but it’s still being peddled as if it does.”

The BlueLeaks documents contain numerous flyers for trainings offered to police agencies across the country. Many of them promote methods of interviewing and interrogation, lie detection, and detecting “danger,” such as Ellory’s, that rest on unsteady scientific ground and have been linked to false confessions and wrongful convictions. The documents offer a window into how various training methods perpetuate myths — subjective, hunch-based approaches to interpreting human behavior that are unreliable and have been discredited by leading psychologists — that police are then encouraged to use in crime solving.

The documents offer a window into how various training methods perpetuate myths that police are then encouraged to use in crime solving.

The search for a foolproof method of lie detection has a “long history,” said Richard Leo, a professor of law and psychology at the University of San Francisco School of Law and an expert on interrogation practices. “The search for some way to be able to read body language, demeanor, vocal pitch, gestures and then infer with a high degree of accuracy whether someone is telling the truth.” It just doesn’t exist, he said. He likens many of the claims about human lie detection to claims of psychic ability. “This reminds me of psychics and the lottery. If there was a psychic and they could see what the lottery numbers are, that would just be gold, right? Why wouldn’t they win $400 million when the Powerball is up there?”

As the country has become increasingly focused on police reform in the wake of George Floyd’s killing by Minneapolis cops, experts say the movement should include reforms to the way police are trained to interview and interrogate suspects, witnesses, and victims to ensure they’re grounded in best practices supported by science. “Part of the distrust that you see between law enforcement and minority communities stems from the way that suspects, witnesses, victims, and family members are treated by detectives during the course of an investigation,” said Steven Drizin, co-director of the Center on Wrongful Convictions at Northwestern University’s Pritzker School of Law, who studies false confessions. Law enforcement training that isn’t based in science “just furthers the deterioration of the relationship between case officers and people in the community.”

Dumb Luck

In addition to Wisconsin’s HIDTA, police agencies in California, Georgia, Nevada, and Texas have promoted Ellory’s trainings, according to flyers found within the BlueLeaks files. One flyer boasts that Ellory has trained law enforcement in the “largest U.S. cities,” including “New York City, Los Angeles, Chicago, Houston, San Antonio, Washington, D.C., Atlanta, Honolulu, Las Vegas, Reno, Key West — just to name a few.” In an email to The Intercept, Ellory said she has been training as Eyes for Lies since 2009 and estimates she’s reached between 2,500 and 3,000 law enforcement officers.

The problem is that what she’s teaching them has been widely discredited — an assertion Ellory vehemently denies. According to Ellory, she was one of 50 individuals identified as an “expert in deception” as part of the so-called Wizards Project, run by researchers associated with Paul Ekman, a professor at the University of California, San Francisco. The researchers studied thousands of people — from CIA and Secret Service agents to regular folks — to see who could best detect behavior associated with deception, a practice that relies heavily on the idea of universal facial expressions and so-called microexpressions that last mere fractions of a second. Ellory’s trainings rely on the validity of both concepts.

While the theory of universal expressions dates back to Charles Darwin, research has been mixed, and Ekman’s work in this area has been repeatedly challenged by scientists in recent years as unreliable, in part because of methodological issues.

Where microexpressions are concerned — also an area of Ekman’s studies — subsequent research has found them “rare and nondiagnostic,” Kukucka said, and that training individuals to see them doesn’t actually work.

Ultimately, Kukucka said, the individuals Ekman identified as exceptional human lie detectors were simply a result of chance. With the Wizards Project, the idea was to test thousands of people to identify those who scored “unusually” high on a lie detection test, Kukucka said. Out of 15,000 people, “they found 50 who were unusually good. And they thought maybe from those people’s knowledge they could reverse engineer — OK, well, what are these people doing that’s working? And then use that to figure out what actually works,” he explained. “The problem with that is, it’s a total artifact of just having a bunch of people and how probability works. If you flip 15,000 coins 10 times, you’re going to get a couple that come up heads all 10 times, but there’s really nothing different about those coins than any of the other coins, just dumb luck.”

Indeed, years of research has demonstrated that behavioral cues — like eye-blinking, arm-crossing, a voice rising or dropping in pitch — are simply not reliable indicators of deception. “A lot of ‘police science’ is really pseudoscience,” Drizin said. “Police officers do believe that they’re able to detect liars from truth-tellers at much higher rates that you and I are. And that’s just been proven not to be the case.” In fact, research has found that the odds of a person detecting deception in another are really no better than chance, and that while those who’ve been trained to do so feel more confident in their conclusions, they’re no more competent. “When police are trained in this false and misleading stuff, they become more confident, so they become more prone to error,” said Leo. “It’s just this loop, this dangerous loop.”

In an email exchange, Ellory first wrote that she wouldn’t have time to explain things to me unless I took one of her courses — her “master class” is currently priced at $1,950 per person — but then noted that she’s not “actively doing” classes right now.

In a subsequent email, she defended her trainings as being rooted in science but wrote that as a “rare expert,” she’s used to people not understanding that. “I find at times with my gift, it’s akin to seeing color in a world where other people live in a colorblind world. Seeing color is ‘real’ but trying to convince a color blind person color exists is nearly impossible,” she wrote. “I tell people in my classes what I teach will be common knowledge in 100 years, but we are still in the dark ages when it comes to understanding human behavior and deception,” she continued. “At a point, I learned, I can’t change the world alone. But I can educate those who are open to learning and they have thanked me endlessly.”

When asked whether it is appropriate to be training law enforcement officers who have power over individual liberty to use scientifically unproven techniques, Ellory retorted that she was “scientifically validated” by Ekman’s research. “I don’t need to reprove it to anyone.”

“You are saying that I shouldn’t teach because I can’t make people like me? Does that mean that Nobel prize winners, acclaimed scientists and researchers who achieve great things shouldn’t teach other people because other people may not reach the same success?” she asked. “Like Lance Armstrong should never coach because he could sustain a heart of 32 beats per minute and consume freakishly low oxygen, but others can’t — so it’s useless?”

“I don’t get that reasoning on any level,” she wrote. “I have insight into human behavior that most people have never considered, don’t understand and when I share it with them through demonstration and example it changes their world for the better. I don’t teach interrogation techniques. I teach people how to seek and find the truth.”

Kukucka called Ellory’s response bizarre. “They’re selling snake oil. I mean, let’s be honest,” he said. “They’re raking in money by selling snake oil to, unfortunately, people who have a lot of clout.”

Ellory’s is not the only training program found among the BlueLeaks documents that sells questionable science to law enforcement. There’s a California-based group that has provided training in neurolinguistic programming, which teaches that deception can be detected by tracking eye movements, a theory that has been widely discredited. And there’s a suite of programs from the Subconscious Communication Training Institute and Spotting Lies, outfits headed by Steven Rhoads — a former police chief, current sheriff’s department investigator, and retired Christian rodeo clown.

The leaked documents indicate that Rhoads’s group has provided a number of trainings over the last decade for law enforcement across the country, including individuals from the Department of Homeland Security and Immigration and Customs Enforcement. The trainings feature lessons in how body language — including “facial gestures and human emotions,” “eye movement and gaze behavior,” and “gestures involving the torso” — can be used in interrogations and reveals not only deception but danger for officers. “As a very general rule of thumb the left side of the body is more apt to reveal known deception than is the right side of the body,” reads material for a 2018 training called “Subconscious Communication for Detecting Danger,” found in files connected to the Northern California Regional Intelligence Center.

Leo says the subconscious communication training is disturbing. “I mean, anything can be said to be subconscious,” he said. “So the cops can just make it up. It’s not based on any research.”

And Kukucka finds the documents related to detecting danger particularly troubling. “I would be very concerned that the context of those trainings would just exacerbate the implicit, especially racial, biases that already exist,” he said. “We know from very clever shooting simulation research that people already hold an implicit bias where their reaction time in shooting unarmed black individuals is faster. So I would wonder from a training like that, the cues that they’re teaching people to look for are those same cues perceived as threatening in black individuals and not in white individuals, for example.”

Rhoads says he’s been teaching interrogation techniques for more than four decades. “I’m still a police officer and use it regularly,” he said. “The techniques I’m teaching work extremely well.” That includes focusing on behavioral cues — including eye movements, like the ones used in neurolinguistic programming. Rhoads, who has a doctorate in behavioral science, said he was one of the “original researchers” into eye movements back in the ’70s, which he’s been able to prove are “98 percent accurate in determining deception.”

But he agrees that researchers are correct to say that you can’t just go into an interview and immediately rely on nonverbal cues to determine deception. Rather, he said, you first have to establish a “baseline” for a person before you can infer deception from their behavior or speech. He says he can do this with high accuracy, usually after asking no more than 20 questions.

Rhoads dismissed the idea that things in a person’s life that an interrogator wouldn’t know — like their cultural norms or past interactions with law enforcement — might influence their behavior during an interview or interrogation. He said his approach for establishing a baseline is similar to what a polygrapher does by measuring physiological responses. “It’s the same science that the polygraph is based on except this is strictly based on verbal and nonverbal leakage versus physiological factors.” Of course, polygraph results are generally inadmissible in court precisely because they’re unreliable.

The approaches that seem to work better to determine whether a person is being deceptive, Kukucka said, “are the ones where the interviewer takes the initiative to be an active participant in the interview and questions a person in a way that draws out things that are diagnostic.” Kukucka said he’d love to see the research that demonstrates Rhoads’s claim of over 90 percent accuracy with his techniques, which he says is just “astronomically higher than anything that any study has ever found.”

In the end, he said, resolving the conflicting claims between trainers like Rhoads and Ellory and researchers like himself should be easy. “If you can do this, prove it. That’s really what it boils down to,” he said. “If you can get 98 percent accuracy with whatever technique you’re using, and you can prove to the scientific community that you can actually do this: A, people are going to throw money at you, and B, we will all gladly be the first to say, ‘You know what? We were wrong, you were right.’”

Resistance Is Futile

Although there are dozens of documents related to deception detection and interrogation trainings by Ellory, Rhoads, and others, the single largest number of documents on the topic that The Intercept identified are for trainings by John E. Reid and Associates, purveyors of the so-called Reid technique. Essentially the granddaddy of interrogation methods, the Reid technique replaced the third degree, and while it does not employ physical torture, it is nonetheless controversial in its approach, which scholars agree has led to false confessions — a persistent problem in the criminal justice system. Roughly 12 percent of the 2,654 exonerations since 1989 involved a false confession, according to the National Registry of Exonerations. Of those who were wrongly convicted of murder and later cleared by DNA, 62 percent had confessed, reports the Innocence Project.

The Reid technique is guilt-presumptive, confrontational, and includes an emphasis on nonverbal behaviors. “It begins with an accusation, a confrontation,” Drizin said. “The police officers have conducted an investigation, and there’s no question in their mind that you were the person who committed the crime.” Interrogators will often lie about evidence linking a person to the crime. “Props are used: big, thick files filled with paper. Claims of DNA or other evidence. Every time the suspect asserts their innocence … they are interrupted and redirected to the idea that they’re guilty.” Reid interrogators also use “themes,” minimizing the crime in a way that makes a confession more likely — offering sympathy, downplaying the severity of the offense, or offering an excuse for it, like you didn’t know what you were doing because you were drunk. It’s “a justification or an excuse that operates as an implied promise of leniency,” said Drizin. “Over time, the message that resistance is futile begins to carry more weight. And then the suspects will agree to confess.”

“Over time, the message that resistance is futile begins to carry more weight. And then the suspects will agree to confess.”

Joseph Buckley, president of Reid and Associates, takes exception to the criticisms heaped on the technique by academics and lawyers and insists that it is supported by science. In response to a series of emailed questions, Buckley directed me to the company’s YouTube channel and a paper he wrote that seeks to clarify what the company calls “misrepresentations” about the practice, though many of them read like distinctions without a difference.

Consider the clarification regarding nonverbal cues. Like Rhoads, Buckley says they shouldn’t be used on their own as an indication of deception, only in context. He offers an example. Say a suspect is asked if he’s ever stolen from his employer. Yes, the suspect says, as he “crosses his legs, looks down at the floor, and dusts his shirt sleeve,” a couple years ago he stole from the hardware store where he worked. But what if a suspect is asked directly, did you steal that missing $2,500? His response — as he crosses his legs, looks at the floor, and dusts his sleeve: “No, I did not.”

“These two subjects displayed identical paralinguistic and nonverbal behaviors during their responses,” Buckley wrote. “However, the interpretation of the behaviors is completely different.” In the first example, the guy is “telling the truth, but he feels embarrassed and possibly even threatened in revealing his prior theft.” But in the second example, the “verbal content … does not explain the accompanying nonverbal behaviors, so the investigator should consider these behaviors as reflecting possible fear or conflict — emotional states that would not be considered appropriate from a truthful subject.”

A 2018 flyer for a four-day Reid training in Austin, Texas, specifically talks about teaching investigators to read behavioral cues — “the verbal and nonverbal behavior symptoms that are displayed by a person who is telling the truth during a non-accusatory interview, as well as those displayed by a person who is withholding or fabricating relevant information,” including “posture changes,” “grooming,” and “eye contact.” On days three and four, the training covers the interrogation process, “beginning with how to initiate the confrontation; develop the interrogational theme; stop denials; overcome objections” and ask questions to “stimulate the admission.”

Building Rapport

Though Reid still dominates the market, there are encouraging signs that may be changing. In 2017, the police-training equivalent of a bomb dropped when Wicklander-Zulawski & Associates, one of the country’s leading law enforcement training organizations, announced that it would no longer teach the Reid technique because of the risk of false confessions. “Confrontation is not an effective way of getting truthful information,” Shane Sturman, the company’s president and CEO, told the Marshall Project. “This was a big move for us, but it’s a decision that’s been coming for quite some time. More and more of our law enforcement clients have asked us to remove it from their training based on all the academic research showing other interrogation styles to be much less risky.”

While science doesn’t support the efficacy of subconscious communication techniques, lie detection, or even the Reid technique, there is ample research to support a different approach: one that is decidedly nonconfrontational, encourages open conversation, and emphasizes rapport-building. Support for this approach in the U.S. comes in part through the work of the High-Value Detainee Interrogation Group, a federally funded interagency effort created by the Obama administration as a means of “advancing the science and practice of interrogation” — and to end Bush-era torture practices against terrorism suspects. The group, known as the HIG, also funded research to develop the science of police interrogation. “Empirical observations have found that police in the U.S. regularly employ poor interview techniques that either reduce the amount of information elicited or entice subjects … to provide incorrect information,” reads a 2016 HIG report. (The HIG was basically abandoned by the Trump administration.)

Increased research in the field, including what has come out of the HIG, “has been paramount,” said Dave Thompson, partner and vice president of operations at Wicklander-Zulawski. Actually, he says, the research has always been there, it just hasn’t always been embraced by practitioners. “We’ve got a lot of these companies that are teaching police practices, regardless of what they are, but they’re teaching it off of being police officers for 30 years. And then you have a lot of academics who are running studies and coming up with great research results but have never been in a practitioner environment. So, I think the really important revolution we’ve had the last few years is the practitioner and the academic working together to make sure that we’re applying research in a practical setting.” Thompson says that’s what his company is trying to do in moving away from what he calls “traditional” interrogation methods.

If the practices that research finds are effective aren’t being incorporated “into what law enforcement’s being trained,” said Thompson, “then we’re headed in the wrong direction.”

There is some suggestion within the BlueLeaks files that newer methods of interrogation might be seeping in, albeit slowly. The documents include at least one flyer from the Savage Training Group advertising “a modern way of interviewing suspects, victims and witnesses that is highly effective and in harmony with the latest research.” The training was organized by the San Mateo County, California, Sheriff’s Office in March. “You might have heard those ‘old-school’ interview techniques have been shown to cause false confessions (Yikes!),” it reads. “You’ve probably been frustrated and thought there ought to be a better way. Well, now there is.”