(Reprinted from The Washington Post)
by Ben Finley
WASHINGTON – Polygraph tests are intrinsically unreliable when used to screen tens of thousands of federal employees for security breaches, a National Research Council panel concluded in a study Tuesday.
And there are no alternatives to the tests, according to the panelists, who examined the polygraph’s utility for the Department of Energy, which operates federal nuclear and weapons research facilities.
“This is a problem for national security and clearly a problem for the DOE,” said Kevin Murphy, a panelist from Pennsylvania State University.
Thousands of federal employees at the Department of Energy, the FBI, the CIA, the National Security Agency and other government facilities face regular polygraph screening.
The study involved only the polygraph’s use as a general screening device for large populations. When focused on a specific crime or unique facts or events, it probably is more reliable, panelists said.
The problem, they said, is that polygraph tests, which measure and record distinctive variations in cardiovascular, respiratory and skin-surface conductivity that suggest deception, produce a lot of false –positive results.
That’s because test responses viewed as deceptive can have other causes, such as stress and exertions, said Stephen Feinberg, a professor of statistics and computer science at Carnegie Mellon University in Pittsburgh who shared the polygraph investigation panel. It was formed by the National Academies of Science, a Washington-based non-profit agency that advises the government on science and technology.
“The research base shows only limited correspondence between deception and the physiological responses monitored by the polygraph,” Feinberg said. This imprecision, he continued, means many individuals are found to be deceptive when they are not.
Arguments against polygraph testing are not new, said Frank Horvath, a spokesman for the American Polygraph Association, based in Chattanooga, Tennessee. He said polygraph tests should not be used alone, but rather along with other deception-detecting methods, including background checks and psychological tests.
The committee estimated that polygraph testing sensitive enough to detect eight out of ten spies in a group of 10,000 federal employees would find 1,600 innocent people guilty and fail to detect two spies. If the polygraphs were made less sensitive, so they would falsely flag fewer than 40 people, eight spies would go undetected, the report found.
Panelist Karen Laskey, an associate professor of engineering at George Mason University in Fairfax, VA observed that the polygraph had never fingered a spy. Among those who beat it, she noted, was Aldrich Ames, a former CIA agent and a spy for the Soviet Union and Russia.