THE DARK SCIENCE OF INTERROGATION
How to find out anything from anyone
By Drake Bennett | Feb. 12, 2015

In August 2003, six months after the U.S. invasion of Iraq and four months into the bloody insurgency that followed, Steve Kleinman, an Air Force lieutenant colonel, arrived in the country as part of a special operations task force based out of Baghdad International Airport. A lean man with an angular face and a faintly Californian cadence, Kleinman had been an intelligence officer for almost two decades. He had questioned high-level prisoners of war during the 1989 invasion of Panama and Iraqi generals during Operation Desert Storm, and he’d run the Air Force Combat Interrogation Course. At the Baghdad airport, however, he witnessed techniques he hadn’t seen in the field. In one of the plywood-walled interrogation rooms he saw a detainee slapped in the face each time he answered a question. Outside another room was a taped-up sheet of paper with the words “1 hour sleep, 3 hrs. awake, ½ hr. on knees, ½ sitting down, 1 hr. standing, ½ hr. knees” written on it. At the bottom it read, “Repeat.”

“This was a year before Abu Ghraib. It was going on in Iraq and Afghanistan, and of course at Guantanamo,” Kleinman says. “Sometimes I got to the point where I had to literally order them to stop. Even then there was surprising blowback. People thought I was coddling terrorists.” Kleinman didn’t think of himself as soft, though, just empirical. In his free time he was an avid consumer of behavioral science research papers, and over the years he’d experimented, in an ad hoc way, with the ideas he found there.

One afternoon a team of Army Rangers brought in a man in his late 30s suspected of selling weapons to the insurgents. By the time Kleinman heard about it, the man had been in custody for three days, enduring hooded stress positions and harsh interrogations, but maintaining a defiant equanimity. “He had these really dark, penetrating eyes. I remember it was almost disconcerting,” Kleinman says. He decided to take over the interrogation himself, and the two men, seated on folding chairs, spoke for three hours. The arms dealer had two young daughters, and he worried for their safety growing up in a violent city. Kleinman pretended that he, too, had two girls, and talked about his worries for them.

After the man grew comfortable, Kleinman tried a thought experiment with him. What if insurgents shot at American soldiers with some of the Kalashnikovs he had sold them and accidentally killed an Iraqi child? Did the arms dealer deserve any blame for the bloodshed that so concerned him as a father?

For a long moment the man just stared at Kleinman. “I thought, OK, this hasn’t worked at all,” he remembers. Then, to his surprise, the man dropped his head into his hands. He’d never thought about it like that, he said. “He told us where his guns were stored. He told us where the guns of his competitors were stored,” Kleinman says. He told them about his neighbor, a chauffeur for two former Ba’ath party leaders, a fact Kleinman hadn’t even known to ask for.


For the most part, interrogators rely on nothing more than intuition, experience,
and a grab bag
of passed-down methods

In intelligence, as in the most mundane office setting, some of the most valuable information still comes from face-to-face conversations across a table. In police work, a successful interrogation can be the difference between a closed case and a cold one. Yet officers today are taught techniques that have never been tested in a scientific setting. For the most part, interrogators rely on nothing more than intuition, experience, and a grab bag of passed-down methods.

“Most police officers can tell you how many feet per second a bullet travels. They know about ballistics and cavity expansion with a hollow-point round,” says Mark Fallon, a former Naval Criminal Investigative Service special agent who led the investigation into the USS Cole attack and was assistant director of the federal government’s main law enforcement training facility. “What as a community we have not yet embraced as effectively is the behavioral sciences.”

Failures of interrogation can have tragic results. Of the 325 people exonerated in the U.S. by DNA evidence, more than a quarter had confessed under interrogation, falsely admitting to violent, often unspeakable acts. As for the war on terror, the Senate Select Committee on Intelligence report released in December on the CIA’s “enhanced interrogation” program is a long and detailed argument that waterboarding, extreme sleep deprivation, rectal feeding, “wall-slamming,” and other techniques were not just harmful, but also counterproductive. Part of the problem, as a 2006 report by the Office of the Director of National Intelligence put it, was a “shortfall in advanced, research-based interrogation methods.”

Rates of True and False Confessions
Tactic True Confessions False Confessions
None 46% 6%
Deal 72% 14%
Minimization 81% 18%
Minimization + Deal 87% 43%
Source: “Investigating true and false confessions within a novel experimental Paradigm” by russano, meissner, narchet, and kassin, American Pscychological society

A small community of interrogators and behavioral researchers in the U.S. and abroad has set out to change that, developing an alternative set of practical, science-based techniques, many of which run counter to what’s been taught in police academies and outlined in the U.S. Army Field Manual. Five years ago, President Obama created a new intelligence body, the High-Value Detainee Interrogation Group, to handle suspected terrorists; one of its teams questioned Boston Marathon bomber Dzhokhar Tsarnaev after his arrest. The HIG, as it’s called, has funded scores of studies.

Christian Meissner, a psychologist at Iowa State University, coordinates much of HIG’s research. “The goal,” he says, “is to go from theory and science, what we know about human communication and memory, what we know about social influence and developing cooperation and rapport, and to translate that into methods that can be scientifically validated.” Then it’s up to Kleinman, Fallon, and other interested investigators to test the findings in the real world and see what works, what doesn’t, and what might actually backfire.

Not too long ago, as part of a test designed by psychologist Melissa Russano, a young woman in a tank top sat at a table with a look of growing apprehension, hunched protectively over her handbag. A student, she had just taken an exam, and a test administrator was accusing her of cheating: Her answers, he said, matched up with those of another student. The administrator said he had just called the professor running the study and reported that he was not at all happy. “He may consider this cheating, I don’t know,” the man said, with sympathy. “I’m sure you didn’t know it would be such a big problem to be sharing. I probably would have done the same thing if I were in your shoes. ... It would ease my professor up if you were seen to be cooperating.” He slid a piece of paper toward her with a confession written on it.

“I don’t think I should sign it. I didn’t do anything,” said the student. Shaking her head, her face pursed in disgust, she signed. As it turned out, she was innocent.

A decade ago, Russano, a professor at Roger Williams University in Rhode Island, set out to design a study that would replicate the social and emotional dynamics of a real interrogation in the lab, where conditions could be controlled. And where, unlike in the messy world of actual cases, the truthfulness of confessions could be easily evaluated. Her study had subjects take a cognitive ability test in a room with another student. Half the time the second student, who was actually working for Russano, would ask for help. The test subjects knew it was against the rules, but most would willingly share their answers. Later, after the test administrator had ostensibly looked over some of the results, he would come back, say there was a potential issue, and leave the subject to stew alone in a room for five minutes. Then some version of the interaction above, taken from a video of one subject, would unfold.

Russano was interested in testing what have long been the twin poles of interrogation styles: “minimization” and “maximization.” They’re forms of coercion that correspond, roughly, to “good cop, bad cop.” Minimization plays down the significance of the crime and offers potential excuses for it—“you just meant to scare her” or “anyone in your situation would have done the same thing.” Maximization plays it up, confrontationally presenting incriminating evidence and refusing to allow any response except a confession. The two are the most widely used tools in the American police interrogator toolkit. The Army Field Manual, which governs all military interrogations, lists approved maximization methods such as “Emotional Fear-Up” and “Emotional-Pride and Ego-Down.”

Lie Detection Accuracy Rate
College Students 45%-60%
Secret Service 64%
Sheriffs 67%
CIA 73%
source: "detecting true lies: police officers’ ability to detect suspects’ lies" by samantha mann, aldert vrij, and ray bull

Chicago-based John E. Reid & Associates is the biggest interrogation trainer in the world, teaching thousands of police officers, intelligence operatives, and private investigators every year. Its techniques are based on the experience of the company’s founders, interviews with suspects after interrogations, and what would appear to be common sense. “In sex crimes, it is also helpful for the investigator to state that he has heard many persons tell about sexual activities far worse than any the suspect himself may relate,” the Reid manual Criminal Interrogation and Confessions advises. “This will serve to encourage the suspect to admit a particularly ‘shameful’ kind of sexual act.”

These techniques do indeed produce confessions. What Russano found, though, was that those confessions are often unreliable. “Guilty people are more likely to confess” when minimization and maximization are used, she says. “The problem is, so are innocent people.” Minimization alone nearly doubled the number of cheaters who confessed in her studies. But it tripled the number of noncheaters who falsely confessed. The videos of those false confessions make for fascinating viewing. Some are angry, some resigned. One young woman keeps her composure until the test administrator leaves the room with her signed confession, then dissolves into tears.

These techniques are a long way from torture—university researchers aren’t likely to run a waterboarding study. They’re also a long way from the sorts of interrogations that have unfolded in precinct houses and desert prisons; after all, as Joseph Buckley, president of John E. Reid & Associates, points out, compared with actual criminals, college students are easy marks. What Russano’s work shows, however, is that even gentler forms of coercion can implicate the innocent and produce misinformation.

Russano is still running versions of that first interrogation study, changing the script to see how it affects the outcome. In one iteration, she explored whether minimization could be purged of the implicit offer of leniency. She had her interrogators be sympathetic, even flattering—saying things such as, “I am sure you are a good person, and no one wants to be accused of cheating or breaking the rules”—but without playing down the seriousness of the offense or its potential punishment. They got just as many true confessions that way, but far fewer false ones.

hundreds of studies have shown that interrogators would be just as well off flipping a coin

Coercive methods are also hobbled by investigators who, far from having some kind of hard-boiled street sense about guilt and innocence, are not necessarily good at knowing whether they’re interrogating a criminal or a valuable witness. The typical interaction between an interrogator and a suspect begins with a determination of guilt or innocence, what Reid calls the Behavior Analysis Interview. But hundreds of studies have shown that interrogators would be just as well off flipping a coin. It doesn’t matter whether they’re simply observing someone speak or asking them questions—the biggest difference between professional and amateur lie detectors is that professionals are much more confident in their abilities.

For one thing, liars often don’t act the way they do on television and in interrogation training manuals. Reid’s Criminal Interrogation and Confessions lists “scratching any part of the body,” “closed, retreated posture,” and lack of eye contact as possible signs of deception. In 2003 a team led by the psychologist Bella DePaulo, then at the University of Virginia, studied those and other cues, synthesizing all the research studies that had been done to that point. Breaking eye contact and fidgeting didn’t actually correlate with lying, they found, nor for that matter did any of the physical tells they looked at. A lot of the behaviors can be symptoms of nervousness, of course, but even an innocent person under interrogation is likely to be nervous. “The basic assumption that liars will be more affected emotionally, it’s very intuitive, but it’s never been supported empirically,” says Par-Anders Granhag, a psychologist at the University of Gothenburg who has trained the instructors at the Norwegian Police University College.

Granhag is a leading proponent of an alternate approach to lie detection that relies on deception’s mental cost rather than its emotional one. Lying is hard: Functional magnetic resonance imaging scans show heightened activity in the brain regions associated with self-control during lying. A liar has to keep his story straight, he has to monitor the listener’s reaction, and he has to monitor himself to make sure he doesn’t look shifty. As a result, lies tend to be told more carefully than true stories—liars speak more slowly, and they include fewer details.

When the psychologist Aldert Vrij, of the University of Portsmouth, asked English police officers to watch videos of people telling stories and pick which ones seemed to be thinking harder rather than which were lying, the officers got better at spotting deception. An interviewer can also add to the cognitive load of the interviewee. A 2008 study led by Vrij, for example, found that when liars were asked to tell their stories in reverse rather than chronologically, they were significantly more likely to be spotted. Other work has suggested that introducing incriminating evidence incrementally, rather than all at once, also makes a liar more likely to slip up.

Common Errors Made Detecting Lies
1
Examining the wrong cues
cues such as gaze aversion and grooming gestures aren’t reliable
2
overemphasis on nonverbal cues
sterotypical beliefs about behavior can lead to false suspicion
3
the othello error
honest people can be as nervous as liars
4
use of Heuristics
“All liars Look Away” is not a reliable measure
5
neglect of personal variations
people respond differently in different contexts
6
neglect of personal differences
introverts can be more at risk for false detection
7
existing interview methods
over-reliance on a manual can impair lie detection
8
Poor Results
Professionals correctly classified liars 56% of the time with a 50% margin of error
Source: “Pitfalls and opportunities in nonverbal and verbal lie detection” by vrij, granhag, and porter, Association for Psychological Science

Buckley refutes work such as Vrij’s. The compromises required to simulate in a lab study what happens in the real world, he argues, render the findings useless. “When you deal with people who are committing mock crimes, there’s no parallel, because they have nothing at stake,” he says. The Reid method is, he freely admits, built on anecdotal evidence, but in an interaction as complex as an interrogation, that’s all there really is. “There’s no scientific research on it. I mean, it’s impossible to do that,” he says. “Human nature is pretty much human nature.”


But what if interrogators avoided coercion altogether? Laurence Alison, a psychologist at the University of Liverpool, has found evidence for the power of rapport-building in real-world interrogations. Alison got access to 181 interrogations of Islamic and right-wing terrorists by British police and coded them for different approaches. He found that the most productive interrogations in terms of information were those in which interrogators essentially acted like therapists. Investigators in Australia, the Netherlands, New Zealand, Norway, and the United Kingdom are now trained to focus not on extracting admissions of guilt but on gathering evidence. That information can back up an innocent suspect’s story, or it can be used to nail a guilty one. Confessions are not discouraged, but British conviction rates suggest that, to juries, the contortions of suspects facing inconvenient facts can be just as damning.

But what if interrogators made no judgments of any kind about their subjects and didn’t use coercion at all?

Much of the HIG-funded research and training has focused on getting better information, even out of people who are perfectly willing to talk. Memories can be as tricky to access for the subject as for the interrogator. One of the limiting factors, it turns out, is the simple act of asking questions. In the early 1980s a young cognitive psychologist named Ronald Fisher noticed he had a strange knack for helping people find misplaced possessions. A friend would call asking if he had left his wallet at Fisher’s apartment, then while talking remember where he’d put it. Fisher realized that without thinking about it he was jogging their memories, subtly guiding the conversation based on theories of memory he was teaching his students at the University of California at Los Angeles. Along with his colleague Edward Geiselman, he started poring over the research literature and watching hundreds of hours of interviews. In a 1985 paper the two formalized the procedure into what they would call the Cognitive Interview.

One of Fisher and Geiselman’s insights was that a typical question-and-answer format makes an interview less likely to yield new information. “Good interviewers ask very, very few questions,” Fisher says. “If you stop and think about it, it makes sense. Why do you ask a question during an interview? Because the witness is not telling you what you want to know.” The more questions asked, Fisher realized, the less likely a witness was to volunteer information, since he was just waiting for the next question.

To address this, the Cognitive Interview emphasizes open-ended questions: Witnesses are told to report everything they remember, whether or not it feels important. They are asked to close their eyes or draw a sketch of how a room was laid out. Interviewers prompt them by asking that they describe the scene from other vantage points—across the street from where they stood or from the car in front of theirs—and to retell a sequence of events starting at different points or in reverse order. Memory is highly associative, and what for a liar is a cognitive challenge can sharpen recall for someone recounting actual experiences, shifting the context in a way that throws previously unnoticed details into relief. In 2013, Fisher ran a study at the Federal Law Enforcement Training Centers in Glynco, Ga. Instructors there were able to get 80 percent more relevant, accurate information using the Cognitive Interview than using the traditional method they taught their students. Police departments and intelligence agencies have begun to adopt it. HIG is teaching federal investigators the method, and the Air Force Office of Special Investigations, the branch’s law enforcement and intelligence arm, is training all of its investigators in it.

Researchers are even moving beyond conversation entirely. In 2012, Maria Hartwig started looking at the role that environment played in investigative interviewing. Hartwig, 34, is a former student of the University of Gothenburg’s Granhag and a professor at the John Jay College of Criminal Justice in New York, where she keeps a small Pinocchio figurine in her office. The basis for her work is a theory called “embodied cognition,” which holds that because our brains first evolved to move our bodies through the world—and only later became capable of higher-level thinking—body and brain still influence each other in elemental, unconscious ways. Some of the best-known work has found that being physically warm makes people more generous toward others and that being excluded socially can lower someone’s estimate of the room temperature.

If the embodied cognition literature was correct, Hartwig reasoned, the typical interrogation room—claustrophobic, locked, austere—was exactly the wrong sort of space to get someone to divulge information. In a yet unpublished study, she redesigned the space around the theme of openness: open windows, an open book on the table, open desk drawers, “a picture of open water under an open sky,” as the paper describes it. She found that subjects provided more detail when questioned in the redesigned interrogation room. The Philadelphia Police Department has expressed interest in the idea, and Hartwig and her collaborators are trying to figure out how exactly they will redecorate interrogation rooms.

Hartwig admits that the work is preliminary. Even if the results hold up, she’s not claiming that the right décor will get terrorists to reveal bomb plots right away, but most interrogators agree that nothing will. “Interrogation doesn’t have a magic switch,” says Colonel Kleinman. “There’s not some ‘break in case of fire’ type of option.” Believing there is, he argues, is part of how you end up with a torture program. In a way, the new tactics are more about doubt than certainty. As deception researchers like to put it, there is no Pinocchio’s nose.

Drake Bennett is a staff writer for Bloomberg Businessweek in New York. He can be reached on Twitter @drakepbennett or by email at dbennett35@bloomberg.net.

Editor: Bryant Urstadt
Design and code: Stephanie Davidson
Photographer: Steven Brahms/Bloomberg Businessweek
Set Design: Jason Singleton