Good Science or Just Bunk?

Controversy surrounds voice analyzers.

by / June 2, 2006
Law enforcement and corrections officers have long used the polygraph as a tool to confront a suspect or prisoner in an investigation. The polygraph is inadmissible in courts, but is widely used to exonerate innocent subjects and elicit confessions from others.

A new voice-analyzing tool -- Layered Voice Analysis (LVA) -- is gaining popularity among law enforcement and corrections agencies.

According to a segment on ABC's Primetime on March 31, 2006, at least 1,500 police or corrections agencies are using some type of voice-stress-analysis software when questioning subjects.

The broadcast prompted the reaction of critics who say the technology doesn't work.

Investigative Tool
The Wisconsin Department of Corrections (DOC) used LVA approximately 125 times in the last two years, according to Dan Westfield, security chief at the DOC. The software is used when questioning inmates during prison investigations. Westfield said LVA is not used to determine guilt or innocence, but rather to steer an investigation in the right direction.

The software measures frequencies in speech by running algorithms and recognizing, with a mathematical formula, patterns indicative of confusion and cognitive dissonance within speech, according to Dave Watson, chief operating officer of V, LLC, which markets LVA.

The software is typically loaded onto a laptop. During the interrogation, the subject talks into a microphone and the words are translated onscreen into graphs that measure the level of confusion or cognitive dissonance in the suspect's voice.

Pop-up boxes alert the interviewer of the deception level detected in the subject's voice with comments, such as: unsure, inaccuracy, probably false or false statement.

"It analyzes the voice and alerts the interviewer that there's something here," Westfield said. "Let's start questioning a little deeper on what we're talking about right now. It helps the interviewer get to the heart of something a little bit quicker."

The DOC doesn't use the polygraph because bringing in a polygraph expert from outside the prison is inconvenient, he said.

"To have our own people trained in use of this technology is more cost effective," Westfield said. "We can investigate and get it resolved quicker, because we have the people with expertise. When you take somebody from outside the DOC, you have to brief them on what's going on. They're not familiar with the environment. It's not like having somebody who works in the prisons; understands the prisons; understands the inmates and what's going on."

Five officers in the DOC are trained to use the LVA software, and were trained for three weeks by V, LLC at a cost of $2,000 each. The software itself costs about $10,000.

The DOC uses the tool to investigate inmate security issues, including suspicion of possession of intoxicants and threats against other inmates and guards, according to Westfield.

"The technology has also been used to clear misconduct where inmates have alleged that another [inmate] did this or that," Westfield said. "So it helped us confirm the fact that an inmate may or may not have done something. It's worked both ways."

In one case, an inmate said he saw another inmate with a shank -- a homemade prison weapon. The test administrator said, based on the results, he believed the accused inmate had in fact possessed an illegal weapon. The inmate later confessed to having the shank and flushing it down the toilet.

As with any tool, LVA is only as good as its operator, proponents say. Formulating questions and good follow-up questions is critical.

"It's like any other tool," Westfield said. "If you don't know how to use the tool, if you don't have some expertise in investigations, LVA is going to do nothing for you."

Truth or Deception?
Critics, however, say this technology doesn't work -- period.

"I think [the systems] have no value as a tool in detecting truth or deception, and that their primary use is to coerce confessions from suspects, both guilty suspects and innocent suspects," said Steve Drizen, legal director at the Northwestern School of Law's Center on Wrongful Convictions, in discussing the software. "They don't work. There have been extensive studies done by government agencies, including the Pentagon, that show that their accuracy in detecting truth or deception is no better than chance.

"Unfortunately law enforcement officers have been willing to pay the $10,000 it costs because they either believe that it's accurate in detecting truth or simply want another expensive toy to use to assist them in obtaining confessions." Drizen continued.

According to the segment on Primetime, the Pentagon used a similar technology in Guantanamo Bay when questioning detainees suspected of terrorism and had subsequently released several of them. The Pentagon later concluded that the software -- the Computer Voice Stress Analyzer (CVSA) -- was unreliable and the federal government ceased using it.

"The Department of Defense did actual scientific research to see if this thing works and they said the accuracy rate was less than the flip of a coin," said Dan Sosnowski, a member of the board of directors for the American Polygraph Association. "Yet we've got agencies using this tool to determine whether people are going to jail, whether people are going to get jobs and that's the biggest problem: Police agencies using it for screening for hiring."

Watson and Westfield said the technology used by the Pentagon was different from LVA, which measures frequencies, not stress.

"It looks at frequencies in the human voice and runs a number of different algorithms that look at those frequencies that correspond to different emotional and psychological states of the speaker," Watson said. "For that reason, it's not designed to try to get confessions or render the ultimate judgment if someone's lying or telling the truth. We call it an investigative focus tool."

Drizen said there are no studies confirming the technology works, and that law enforcement does use it to force confessions.

The most noteworthy case is that of 14-year-old Michael Crowe, who was extensively interrogated by California police and eventually confessed to the murder of his 12-year-old sister, when he was, in fact, not involved. Crowe's story is well documented in national media outlets.

Another case involved a 12-year-old boy who was told by police in New Philadelphia, Ohio, that there was evidence he murdered a 5-year-old girl and was threatened that the analyzer would catch him in a lie, Drizen said, adding that the boy was scared of the analyzer and confessed, even though he was innocent.

"Most false confessions involve situations where suspects are persuaded that it's in their best interest to confess, and that's done through a number of tactics," Drizen said. "One of them is to deceive the suspect by telling the suspect that the evidence against them is rock solid. Then investigators motivate the suspect to confess by suggesting to them that confession will bring benefits to them in the court system like a lesser sentence or less significant charge."

Drizen said there's no difference between LVA and CVSA systems.

"I don't think any is more scientific than the other. Both are equally unscientific as far as I'm concerned," he said. "The problem with these tests is quite simply that differences in heart rate or differences in voice stress can be the result of stress just as easily as it can be the result of deception."
Jim McKay, Justice and Public Safety Editor Justice and Public Safety Editor