© John J. Kim/MCTDonny McGee was arrested for murder in 2001 after a Chicago police officer and a detective fabricated a polygraph result and false confession and spent three years in prison before being acquitted in a jury trial in 2004. McGee plays with his children in Bourbonnais, Illinois.
As a result, innocent people might have been labeled criminal suspects, faced greater scrutiny while on probation or lost out on jobs. Or, just as alarming, spies and criminals may have escaped detection.
The technical glitch produced errors in the computerized measurements of sweat in one of the most popular polygraphs, the LX4000. Although polygraphers first noticed the problem a decade ago, many government agencies hadn't known about the risk of inaccurate measurements until McClatchy recently raised questions about it.
The manufacturer, Lafayette Instrument Co. Inc., described the phenomenon as "occasional" and "minor," but it couldn't say exactly how often it occurs. Even after one federal agency became concerned and stopped using the measurement and a veteran polygrapher at another witnessed it repeatedly change test results, the extent and the source of the problem weren't independently studied nor openly debated. In the meantime, tens of thousands of Americans were polygraphed on the LX4000.
The controversy casts new doubt on the reliability and usefulness of polygraphs, which are popularly known as lie detectors and whose tests are banned for use as evidence by most U.S. courts. Scientists have long questioned whether polygraphers can accurately identify liars by interpreting measurements of blood pressure, sweat activity and respiration. But polygraphers themselves say they rely on the measurements to be accurate for their daily, high-stakes decisions about people's lives.
"We're talking about using a procedure that has a very weak scientific foundation and making it worse," said William Iacono, a University of Minnesota psychology professor who's researched polygraph testing. "I already don't have very much confidence in how government agencies conduct these tests. Now, they might as well be flipping a coin."
Despite the scientific skepticism, intelligence and law enforcement agencies see polygraph as useful in obtaining confessions to wrongdoing that wouldn't otherwise be uncovered. Fifteen federal agencies and many police departments across the country rely on polygraph testing to help make hiring or firing decisions. Sex offenders and other felons often undergo testing to comply with probation or court-ordered psychological treatment. Police detectives and prosecutors rule out criminal suspects who pass and scrutinize those who don't.
In its ongoing series about polygraph use by government agencies, McClatchy found that such testing has flourished despite being banned for use by most private employers 25 years ago. For federal jobs alone, more than 70,000 people are polygraphed each year, and most can't challenge the results in court or allege abusive tactics. While supporters say accuracy can be 85 to 95 percent, polygraphs aren't required to meet any independent testing standards to verify the accuracy of their measurements, unlike medical or other computerized equipment.
The concerns about the LX4000 only add to the criticism.
"If you buy all of the propositions that the physiological measurements are a reliable proxy for truth telling or deception, then the whole premise depends upon a machine that can precisely record those measurements," said Gene Iredale, a San Diego attorney. "If you don't have that, then you have a hope piled on a speculation, and on top of it all an error-filled system."
In the absence of an independent assessment, polygraphers depend on the federal government, the manufacturer or one another to be notified of a problem with the technology. Many polygraphers, however, told McClatchy they didn't know about the possibility of inaccurate measurements or that they could occur in other polygraphs that use the same technology.
"If this was being debated, I would have liked to have known about it," said Danny Fields, the supervisor of the Kansas City Missouri Police Department's polygraph section, which used the LX4000 for years before replacing it with a newer Lafayette model. "I believe in polygraph 100 percent, but I want to make sure it's working like it's supposed to be working."
© Melina Yingling, Judy Treible, Danny Dougherty What a polygraph measures: Sensors measure body changes as subject answers yes-no questions; data recorded on computer.
The polygraph profession is highly secretive, and many agencies cite national security or law enforcement interests as barring them from answering questions. Of 63 federal, local and state agencies contacted by McClatchy, only 40 would say what types of machines they used. Of those, 27 state and local agencies said they either currently use or have used the LX4000, including big city police departments such as New York, Chicago and Los Angeles. None of those agencies said they were reviewing their previous tests for errors, although some people who'd failed polygraph tests said they thought they'd been inaccurately labeled. Ten federal agencies have used Lafayette polygraphs.
"I'm astounded that a government agency would rely on this machine to make any decision," said John Stauffer, a Chicago accountant who was denied an FBI job in 2011 because he didn't pass his polygraph test. "I've always known that I shouldn't have failed. Now I wonder whether this was the problem."
Scientists have experimented for more than a century with running a minuscule amount of electricity through sweat glands in the fingertips as a way to gauge emotions and mental effort. In the past two decades, however, polygraphs marketed to government agencies have changed the way perspiration is measured.
As a result, the LX4000 measures sweat in two ways. One method, known as the manual mode, directly measures the secretions from sweat glands, as scientists traditionally have done. The other, known as the automatic mode, electronically filters the measurements and is designed to smooth out the sometimes erratic graphic representations and make them easier to interpret.
David Reisinger, a veteran federal polygrapher, said he first witnessed a problem with the LX4000 in 2005, while discussing a test with a Lafayette employee by phone. When he switched between the two modes, he noticed a difference in the measurements.
"It was so significant I noticed the problem immediately," said Reisinger, a polygrapher at the Defense Intelligence Agency at the time. "It jumped right off the screen at me."
Reisinger pressed the company to look into it because he saw it could change the outcome of a test depending on the setting. Polygraphers assign numbers to sweat measurements and add them up for a final score that's supposed to show whether someone is lying. In a test where one point can make a difference, Reisinger documented up to a 16-point difference between the two modes.
He notified his supervisors, and Lafayette pledged to fix it. Years and dozens of examples later, the company still hadn't, he said.
"What troubled me is that they couldn't tell me which measurement was accurate," he said.
The Air Force's Office of Special Investigations noticed a problem as early as 2002, the year the LX4000 hit the market. A spokeswoman said her law enforcement agency was concerned that it could change the outcome of tests, and sought out Lafayette officials.
"They recommended that all polygraph charts be collected in the manual mode," spokeswoman Linda Card said in a statement. "As a result of the possible flaw in the automatic mode, we, as an organization, directed the use of only the manual mode" during testing.
The manufacturer's advice apparently didn't reach other government agencies.
"This is news to me," Reisinger said. "The manufacturer never told us that."
Instead, Lafayette described the problem to the Defense Intelligence Agency as "minor" and repairable. In one email , then -Lafayette Operations Manager Mark Lane told the agency in 2007 that company officials felt "extremely confident" they could fix it, adding that they had "devoted our entire engineering efforts" to fixing the automatic mode.
"We certainly agree that this is an issue that needs to be resolved," he told DIA officials.
Word of the problem spread unofficially. At one point, a CIA employee called Reisinger to request information about the issue. He sent along what he knew.
In 2009, Reisinger told his bosses that he'd seen several software updates and none of them appeared to fix the problem. By then, he'd become an official who reviewed tests to make sure polygraphers were complying with federal standards. He recommended that the Defense Intelligence Agency stop using the LX4000. The DIA, which decided not to switch machines, refused to comment for this story.
"I felt the manufacturer had given us the runaround and was never completely straightforward about what was going on," Reisinger said. "I didn't think we could trust them anymore."
According to a 2002 federal training document, Lafayette advised polygraphers to use the automatic mode, although the company now says it doesn't recommend one mode over the other.
"Lafayette Instrument Company . . . has helped customers to select technology and procedures that best serve their objectives," the company said in a statement.
After a McClatchy reporter asked about the problem, Lafayette sent a notice to customers in March acknowledging that a difference in measurements could occur but described it as a "rare" phenomenon that it had attempted to eliminate with improvements to its machines. The company pledged to test its polygraphs to determine the extent of the problem, although it added in a statement to McClatchy that "anecdotal experience tells us that different . . . modes are in agreement most of the time."
Lafayette, meanwhile, has been marketing a new model, the LX5000, that it said has the same "potential for occasional differences," like any polygraph that has an automatic mode.
When McClatchy asked the manufacturer why it hadn't sent out a notice earlier, officials responded that it was "not productive" to discuss such a question.
"In this case, it is impossible to speculate about 'why' because what you perceived as a newly discovered 'problem' is actually a known" phenomenon,Lafayette wrote to a reporter.
Gary Berntson, an Ohio State University professor of psychology who's studied such sweat measurements, agreed with the company that differences in measurements were likely a rare occurrence.
"But the cost of such an outcome to the individual could be huge," said Berntson, a former consultant to Lafayette. "The crux here is proper disclosure. The manufacturer needs to alert the user of potential biases, however subtle or rare."
Charles Honts, a psychology professor with Boise State University who's also researched polygraph, questioned why the manufacturer or the federal government didn't weigh in with clearer guidance since both witnessed the problem repeatedly.
Honts and several other scientists who research polygraph previously warned government agencies not to rely only on the automatic mode because research had shown the risk of errors in that measurement in any polygraph.
"The insidious thing is that this phenomenon biases tests against the innocent, and the government knows that," said Honts, who's worked on research for a Lafayette competitor. "This is just another example of science being ignored."
In fact, Lafayette sent McClatchy documents of what it described as evidence that the same problem occurs in an unnamed competitor's polygraphs, but it asserted "the examples provided . . . are in fact not common whether using our instrument or those (of) our competitors."
One competitor, Limestone Technologies Inc., has made public assertions that its polygraph measurements are better than Lafayette's.
Scientists, however, said it was impossible to know whether one company's polygraph is better than another's without independent testing, adding that the problem with differences between the two modes was likely made worse by bad practices. Many agencies use stainless steel electrodes without the recommended gel for measuring the sweat, which could yield erratic readings in the more reliable manual mode. As a result, polygraphers might be tempted to turn to the automatic mode, which is viewed as easier to interpret.
"We've been saying this for 30 years," said John Kircher, a researcher and psychology professor at the University of Utah who helped invent one of the earliest computerized polygraphs in 1979. "But no one seems to be listening."
The National Center for Credibility Assessment, the federal polygraph training academy, declined to comment, citing national security concerns.
The FBI, Secret Service, Customs and Border Protection, Drug Enforcement Administration, State Department and Bureau of Alcohol, Tobacco, Firearms and Explosives also refused to comment, but federal records show they've contracted to buy Lafayette's polygraphs, often without competitive bidding. According to company and government documents, a new LX5000 can range from $3,200 to $9,500. The CIA and the National Reconnaissance Office also have used the company's polygraphs, but those contracting records are kept secret and officials there wouldn't comment.
Several police departments, however, said they weren't disturbed by the news because they hadn't noticed inaccurate measurements. Tommy Thompson, spokesman for the Phoenix Police Department, said his polygraphers had "over 60 years of experience collectively" and added that "no disparities have ever been noticed." Others said they thought supervisors would catch any error before it had a significant impact on a test.
But Phoenix and many other government agencies generally rely on the automatic mode. As a result, veteran polygraphers would be unlikely to notice any difference between the two measurements, even if it were occurring routinely.
"If you don't flip to the other mode, you'd be oblivious to the fact that it's even different," Reisinger said.
Scientists said the federal government and manufacturers should be warning polygraphers about the technological pitfalls. Iacono conducted his own laboratory tests of older Lafayette machines in the 1990s and didn't find any issue with their earlier measurements. In fact, he found the machines to be as good as scientific equipment that cost much more.
"Because of that, I have never criticized the actual measurements, and I've always thought they've only made them better," Iacono said. "What this shows is that they've actually made them worse and they can't fix the problem."
Did they know that psychopaths lying turn up as telling the truth?