By Matt Stroud | Bloomberg | February 2, 2015
This article features interview with CLBB Co-Director Judith Edersheim, JD, MD. Dr. Edersheim’s 2014 paper, “A Polygraph Primer: What Litigators Need to Know,” written with Ekaterina Pivovarova, Justin Baker and Bruce Price, about the accuracy of the polygraph test, is cited.
By the beginning of February 1935, defense and prosecution attorneys in the criminal trial of Tony Grignano and Cecil Loniello were at a crossroads. The two men stood accused of attempting to kill a sheriff in Portage, Wis., and all of the evidence was circumstantial—the word of the sheriff against the word of the defendants. But Judge Clayton F. Van Pelt thought he could change that.
The judge had heard about the work of the Scientific Crime Detection Laboratory at Northwestern University’s School of Law, where Professor Leonarde Keeler had spent more than a decade tinkering with a portable device to measure the responses of skin and blood pressure during questioning. In the hands of a trained expert, Keeler said, the device could help identify whether someone was telling the truth. While the Keeler Polygraph had been a showpiece at the 1933 Chicago World’s Fair, it had never been used in a criminal trial. On Feb. 2, 1935—exactly 80 years ago—Keeler used his polygraph test and determined that Grignano and Loniello were likely lying. The professor and his machine ended up persuading the jury.
After the conviction, a reporter asked Keeler what it all meant. The professor answered confidently: “It means that the findings of the lie detector are as acceptable in court as fingerprint testimony.”
But eight decades later, polygraphs are not admissible in court. Lie detectors are used for background checks and job screenings—and in police investigations when officers want to extract information from a suspect. But lie detectors aren’t allowed inside the courtroom during criminal cases, and it’s largely because scientists are far from impressed with the Keeler’s modern-day heirs.
A team of doctors from Harvard’s Center for Law, Brain and Behavior pointed out in a paper last year that polygraph studies yield “widely divergent rates of accuracy in detecting deception, some as low as chance and others as high as 95 percent.” The problem is that every polygraph test is different. At its core, a standard polygraph test involves a trained examiner hooking a person up to a polygraph device and asking questions at varying levels of provocation. The device monitors and records how the person’s body responds, and an examiner interprets the output.
Questioning structures such as the Guilty Knowledge Test have sought to standardize the interrogation format, but administering a polygraph hasn’t changed much since 1935. Even the professor himself allowed that the polygraph wasn’t foolproof, telling jurors at the original Wisconsin trial that his test was about 75 percent accurate. Decades later it’s still up to the examiner to ask effective questions and to interpret the output accurately; there are good polygraph examiners and there are bad polygraph examiners.
“The political and legal argument some make in favor of the polygraph is that it’s very accurate depending on who the examiner is,” says Dr. Judith G. Edersheim, co-director of Harvard’s Center for Law, Brain & Behavior. “But for a scientist, saying it’s examiner-dependent means it’s not reliable.”
Skeptical scientists have kept the U.S. judicial system from widely embracing the polygraph. The U.S. Supreme Court took up the question of lie detector admissibility in a 1998 case, United States v. Scheffer. A military court had decided that exclusion of polygraph evidence violated the Sixth Amendment right to present a defense. The high court disagreed: “A fundamental premise of our criminal justice system is that ‘the jury is the lie detector.’” While experts who testify about fingerprints or ballistics can weigh in on matters related to the question of guilt or innocence, a lie detector purports to answer it: Is this person lying or not? Since the polygraph is not consistent or perfect, the high court ruled it out of bounds.
By the time of the Supreme Court’s ruling, there was more than a half-century of questions about the accuracy of lie detection. Back in Keeler’s day, though, the polygraph worked its way into American lore. The inventor of the test was soon administering polygraphs in some of the nation’s most notorious cases. By 1938, he left Northwestern to start the Keeler Institute, which trained polygraph examiners and provided expert testimony in court cases. The Federal Bureau of Investigation would eventually begin conducting polygraph examinations for criminal investigations and job screenings. Keeler even landed a movie role, playing himself in the 1948 James Stewart film, Call Northside 777. Keeler died in 1949, and his lie detector had become a cultural mainstay. Two years later the U.S. Army Polygraph School began training its first class of examiners in Keeler’s method.
Even though the polygraph now has a tarnished legal reputation, modern-day version of the exam are still used in job screenings, background checks, and reality television shows. Contemporary researchers haven’t given up on lie detection technology, and there have been significant advances. The two best-known post-polygraph lie detector exams are the P300 (also known as the Electroencephalography, or EEG), and the functional Magnetic Resonance Imaging (fMRI).
P300 technology centers around the brainwave for which the exam is named, which is activated when someone views a familiar object. The technique involves a subject wearing electrodes and a headband while being shown images on a computer screen, allowing an examiner to determine if the subject has stored information that she shouldn’t.
The fMRI tracks how blood flows to certain areas within the brain. The idea is that the truth is easy; the brain has to work harder to suppress a piece of information and construct a lie. The exam involves sliding a person into an fMRI machine and creating an image of a lie within the brain.
Both next-generation lie detectors have run into problems similar to the polygraph. “Not all lies are created equal, and not all brains are created equal,” Edersheim says. “Does a half-truth look the same as a whole lie? If you believe in your lie deeply, will that look different on a brain scan from a lie you experience as deceptive?” Neither the P300 nor the fMRI can say for sure, so courts have likewise backed away from these tests. “The courts have said, ‘Look guys, if you can’t sort these things out, you’re not ready for prime time,’” Edersheim says.
A 2012 case decided by the Sixth Circuit, U.S. v. Semrau, asked the judges to weigh in on the admissibility of fMRI tests. “[T]here are concerns with not only whether fMRI lie detection of ‘real lies’ has been tested but whether it can be tested,” wrote Judge Jane B. Stranch in her opinion.
Since courts are opposed to admitting lie detection, lie-detection innovators are avoiding the courtroom altogether.
Judee Burgoon and Dr. Jay Nunamaker at the University of Arizona are at work on a screening system called Avatar, which is short for Automated Virtual Agent for Truth Assessments in Real-Time. It’s designed to be a fast, advanced, polygraph-like test that doesn’t involve hooking anyone up to a machine or administering an fMRI. Instead, Avatar is a computer program designed to look like a face on a screen. People reply to the face as it poses quick, direct questions. It’s hooked up to a camera, artificial intelligence, and noninvasive sensors that observes cues: pupil size, thermal indicators, cardiorespiratory indicators, blinks, voice indications, and more. The goal: to determine whether someone is acting suspiciously.
The Avatar project is funded by the Department of Homeland Security and is still in a testing phase. But it’s designed to be used at ports of entry, borders, airports, detention centers, and anywhere else where lots of people need to be screened as quickly as possible. If someone’s found to act suspiciously, an agent on the ground will be summoned to investigate. “This isn’t a lie detector,” Nunamaker says. “We’re looking for signs of risk.” In fact, he says, looking for a foolproof lie detector may be impossible: “Everyone we talk to wants to know, ‘What’s Pinocchio’s nose?’ Well, that’s never gonna happen.”
“Nothing is ever going to 100 percent determine if someone’s lying,” agrees Burgoon. “There’s enough variability with humans that we shouldn’t assume that any machine is going to be able to fully capture what’s going on inside someone’s brain. When people start having those kinds of expectations, they’re more in the sci-fi realm.”
“They’re looking for something that doesn’t exist,” Burgoon says.
Professor Keeler had similar reservations about promising too much with his polygraph test. When asked by the Wisconsin judge 80 years ago if he would allow the test “to decide the most important affairs of your life,” Keeler answered in negative. “I wouldn’t want to convict a man on the grounds of the records alone,” he said.