Category Archives: Voice Stress Analysis

Stress and Deception in Speech: Evaluating Layered Voice Analysis

Hot off the press in Journal of Forensic Sciences (hat tip Mind Hacks), a study in which a Layered Voice Analysis system was tested independently and found to be effective at the chance level. In other words, you might as well flip a coin.

Here’s the abstract:

This study was designed to evaluate commonly used voice stress analyzers—in this case the layered voice analysis (LVA) system. The research protocol involved the use of a speech database containing materials recorded while highly controlled deception and stress levels were systematically varied. Subjects were 24 each males/females (age range 18–63 years) drawn from a diverse population. All held strong views about some issue; they were required to make intense contradictory statements while believing that they would be heard/seen by peers. The LVA system was then evaluated by means of a double blind study using two types of examiners: a pair of scientists trained and certified by the manufacturer in the proper use of the system and two highly experienced LVA instructors provided by this same firm. The results showed that the “true positive” (or hit) rates for all examiners averaged near chance (42–56%) for all conditions, types of materials (e.g., stress vs. unstressed, truth vs. deception), and examiners (scientists vs. manufacturers). Most importantly, the false positive rate was very high, ranging from 40% to 65%. Sensitivity statistics confirmed that the LVA system operated at about chance levels in the detection of truth, deception, and the presence of high and low vocal stress states.

Reference:

You’ll find more on Layered Voice Analysis in the voice analysis category on this blog.

Voodoo science in fMRI and voice analysis to detect deception: compare and contrast

Controversy and debate is the driver of scientific progress.  It forces us to re-examine our assumptions, scrutinise our methods and think hard about the meaning of data.  Of course, there is another way of dealing with controversy…

Voodoo science in fMRI

If you’re involved or simply interested in fMRI research you’ll already be well aware of the ongoing debate about Voodoo Correlations in Social Neuroscience [pdf]. If not, you’ll find the detail in coverage all over the psych and neuroblogs by googling the title or simply “voodoo correlations”.

Here’s how it went:

1. Edward Vul, Christine Harris, Piotr Winkielman, and Harold Pashler wrote a critique of a series of recent research studies exploring the neural correlates of various social psychological issues. Their paper was accepted by a peer-reviewed journal and will be published later this year.

2. Authors of those criticised research papers wrote careful defences of their work and pointed out problems in Vul et al’s arguments (here and here).

3. Vul et al. responded to the criticisms here.

And the debate continues – watching from the sidelines you get a sense of the passion and the intellect on both sides, with the process of open debate resulting in further clarification and some concessions (on both sides). Ultimately, this debate will result in better understanding of some important issues and better scrutiny of new research. Scientific progress, in other words.

Voodoo science in deception detection

Compare this to another recent controversy that started in the research literature (hat tip to Mind Hacks).

1. In 2007, the International Journal of Speech Language and the Law (a peer reviewed journal) published a critique by Anders Eriksson and Francisco Lacerda of mechanical methods of deception detection that claim to use ‘voice stress analysis’ or ‘layered voice analysis’ to detect deception. It is more pointed and more personal than the Vul et al. critique (commenting on the companies and the individuals involved in developing and marketing such machines), but the authors nevertheless examine the scientific literature carefully and raise some significant problems with the technology as it is marketed.

2. One of the companies named, Nemesysco, threatened to sue.

3. The publishers of IJSLL withdrew the paper (though, this being the age of the internet, you can access it here).

Rather than publish the potentially ground-breaking scientific evidence underpinning their technique, respond to the criticisms or engage in debate, a company uses legal threats to silence criticism. The result is that we have no chance to hear both sides of the story, little chance of increasing our understanding of the techniques or their theoretical basis, further polarisation of the pro- and anti- camps, and bugger all scientific progress. Shame.

Of course, Nemesysco’s actions do mean that a lot more of us know and are talking about the criticism of their technology than had they let the journal article lie (no pun intended).

Voice Stress Analysis: Only 15 Percent of Lies About Drug Use Detected in Field Test

The latest issue of the National Institute of Justice journal (NIJ Journal No. 259, March 2008) features a great article by Kelly Damphousse summarising recent research on voice stress analysis (VSA). Here’s an extract:

According to a recent study funded by the National Institute of Justice (NIJ), two of the most popular VSA programs in use by police departments across the country are no better than flipping a coin when it comes to detecting deception regarding recent drug use. The study’s findings also noted, however, that the mere presence of a VSA program during an interrogation may deter a respondent from giving a false answer.

The findings of our study revealed:

  • Deceptive respondents. Fifteen percent who said they had not used drugs—but who, according to their urine tests, had—were correctly identified by the VSA programs as being deceptive.
  • Nondeceptive respondents. Eight and a half percent who were telling the truth—that is, their urine tests were consistent with their statements that they had or had not used drugs—were incorrectly classified by the VSA programs as being deceptive.

Using these percentages to determine the overall accuracy rates of the two VSA programs, we found that their ability to accurately detect deception about recent drug use was about 50 percent.

Based solely on these statistics, it seems reasonable to conclude that these VSA programs were not able to detect deception about drug use, at least to a degree that law enforcement professionals would require—particularly when weighed against the financial investment. We did find, however, that arrestees who were questioned using the VSA instruments were less likely to lie about illicit drug use compared to arrestees whose responses were recorded by the interviewer with pen and paper.

Damphousse concludes:

It is important to look at both “hard” and “hidden” costs when deciding whether to purchase or maintain a VSA program. The monetary costs are substantial: it can cost up to $20,000 to purchase LVA. The average cost of CVSA® training and equipment is $11,500. Calculating the current investment nationwide—more than 1,400 police departments currently use CVSA®, according to the manufacturer—the total cost is more than $16 million not including the manpower expense to use it.

The hidden costs are, of course, more difficult to quantify. As VSA programs come under greater scrutiny—due, in part, to reports of false confessions during investigations that used VSA—the overall value of the technology continues to be questioned.

See also:

Computer voice stress analyzer tests debated

Another TV expose of the use of Computer Voice Stress Analyzers in the USA, this time from Colorado’s 9 News (1 November):

A device used by Colorado law enforcement agencies to identify when someone is lying, may not work and may be costing taxpayers money. Computer Voice Stress Analyzers (CVSAs) claim to measure changes in a person’s voice that indicate a lie. However, three recent studies say the device does not accurately tell the difference between a person lying and a person telling the truth. CVSAs have been used by 21 law enforcement agencies in Colorado.

Can lie detectors be trusted?

Detailed commentary from Patrick Barkham in the Guardian (18 Sept), exploring the use of ‘lie detecting’ machines in the UK. He covers the use of voice stress analysis in benefit offices and insurance companies, and polygraphy for sex offenders. Interesting stuff, and well worth reading in full over on the Guardian site. Here’s a flavour:

[Harrow] council prefers the phrase “voice risk analysis” and Capita calls its combination of software, special scripts and training for handlers the “Advanced Validation Solution”. Just don’t say it’s a lie detector. “Please don’t call it that. We’re not happy with that. It’s an assessment,” says Fabio Esposito, Harrow’s assistant benefit manager.

… Voice stress analysis systems have been used for more than five years in the British insurance industry but have yet to really catch on, according to the Association of British Insurers. There was an initial flurry of publicity when motor insurance companies introduced the technology in 2001 but it is still “the exception rather than the norm,” says Malcolm Tarling of the ABI. “Not many companies use it and those that do use it in very controlled circumstances. They never use the results of a voice risk analysis alone because the technology is not infallible.”

… Next year, in a pilot study, the government will introduce a mandatory polygraph for convicted sex offenders in three regions. … Professor Don Grubin, a forensic psychiatrist at Newcastle University… admits he was initially sceptical but argues that polygraphs are a useful tool. “We were less concerned about accuracy per se than with the disclosures and the changes in behaviour it encourages these guys to make,” he says. “It should not be seen as a lie detector but as a truth facilitator. What you find is you get markedly increased disclosures. You don’t get the full story but you get more than you had.”

…critics argue that most kinds of lie-detector studies are lab tests, which can never replicate the high stakes of real lies and tend to test technology on healthy individuals (usually students) of above-average intelligence. Children, criminals, the psychotic, the stupid and even those not speaking in their first language (a common issue with benefit claimants) are rarely involved in studies.

Using voice analysis to detect benefit cheats

ratlieThe media is reporting that a pilot scheme in the UK to use voice stress analysis (or, more accurately, “voice risk analysis”) on benefit applicants is a success. The Observer headline proclaims Technology set to be introduced nationwide after pilot saves £110,000 (2 September):

Benefit claimants and job seekers could be forced to take lie detector tests as early as next year after an early review of a pilot scheme exposed 126 benefit cheats in just three months, saving one local authority £110,000.

The news report also points out that many are skeptical:

Experts in America, where the most comprehensive scrutiny of the technology has taken place, warn that the technology is far from failsafe. David Ashe, chief deputy of the Virginia Board for Professional and Occupational Regulation, said, ‘The experience of being tested, or of claiming a benefit and being told that your voice is being checked for lies, is inherently stressful. ‘Lie detector tests have a tendency to pass people for whom deception is a way of life and fail those who are scrupulously honest.’

Reading beyond the headlines, it’s clear that the pilot study is not finished, it hasn’t been properly evaluated, and no decision has yet been made. In Lie detector beats benefit fraud, silicon.com (3 Sept) reveals

A spokesman for the Department for Work and Pensions (DWP) – which funded the pilot – told silicon.com the department will evaluate the technology when the trial is completed next May. He said the DWP will “look at the evaluation results and see if it’s viable, see if it’s something to work on and see if other councils are interested in doing it”. If the benefits are seen as sufficient, the system could potentially be rolled out across the country, although no firm plans are currently in place.

But this hasn’t stopped others jumping on the VSA bandwagon, as the Telegraph (9 Sept) and BBC Online (7 Sept) report that Birmingham Council is next in line to adopt the system.

More Deception Blog posts on this story here and here, and more generally on VSA here.

Photo credit: niznoz, Creative Commons License

Voice Stress Analysis – a new report (still doesn’t work)

telltruthHat tip to the Anti-Polygraph Blog for alerting us to a new study to test the efficacy of Voice Stress Analysis. From the study’s abstract:

…The goal of this study was to test the validity and reliability of two popular VSA programs (LVA and CVSA) in a “real world” setting. Questions about recent drug use were asked of a random sample of arrestees in a county jail. Their responses and the VSA output were compared to a subsequent urinalysis to determine if the VSA programs could detect deception.

Both VSA programs show poor validity – neither program efficiently determined who was being deceptive about recent drug use. The programs were not able to detect deception at a rate any better than chance. The data also suggest poor reliability for both VSA products when we compared expert and novice interpretations of the output. …

However, the researchers did find that arrestees who knew they were going to be VSA tested “were much less likely to be deceptive about recent drug use than arrestees in a non-VSA research project” (though they do admit that the non-VSA project was not carried out in exactly the same way as the VSA study). The authors suggest that regardless of its validity, a VSA device may produce a bogus pipeline effect, which is perhaps why so many law enforcement agencies believe that it works:

When police officers report that VSA programs “work,” they generally mean that they were able to obtain a confession from suspects by telling them that the computer “said they were lying.” The potential problem, of course, is with false confessions. Several high profile cases have emerged in the past decade that suggest impressionable suspects may confess to a crime that they did not commit because they believed the software. The rationalization is usually that they “must have forgotten” that they did it. Obviously, the bogus pipeline effect of VSA products has important positive and negative implications (p.86).

The authors also highlight the financial implications, estimating the cost for training just one person from each of the 1400 law enforcement agencies who claim to use VSA at more than $16 million. The software and laptop is nearly $10K per agency. And “computer upgrades can increase the cost to almost $13,000″ (p.4). Yikes.

Download the full text report as a pdf from the link below, and read more on the Deception Blog about VSA here and about the bogus pipeline effect here.

Reference

Photo credit: arimoore, Creative Commons License

Deception links from around the web

linksSome quick deception-related links from around the blogosphere:

PsyBlog presents the “Top 3 Myths, Top 5 Proven Factors” on lie detection (12 May).

Wired (10 May) picks up on the UK government trial of voice stress analysis for alleged benefit cheats.

The Psychjourney Podcast for 27 April is on Malingering and PTSD (mp3).

If podcasts are your thing you can also listen to an interview with Ken Alder, author of a new book on the polygraph, on the Bat Segundo show (mp3). As the Anti-Polygraph Blog points you, you have to sit through a little silliness first…

Photo credit: mklingo, Creative Commons License

Don’t make benefits claimants take lie detector tests says TUC

The Trades Union Congress has called for the Department for Work and Pensions to abandon plans to use voice stress analysis in benefit centres (press release, 4 May). Quite right too. They say:

The Government should abandon plans to trial lie detector tests for people claiming benefits because the accuracy of the technology has not been scientifically proven, and individuals with genuine cases are likely to be discouraged from applying for the help they desperately need, says the TUC today (Friday).

[…] a TUC briefing ‘Lies, damned lies and lie detectors’ says that the science just isn’t there to back up the technology, and any use of the software when dealing with benefit claimants means that the innocent are just as likely to fall foul of the system as the genuinely guilty.

The TUC says that the problem with the lie detection technology that the DWP intends to use is that it cannot detect lies. Voice risk analysis and lie detectors can only detect, with varying accuracy, changes in the body, such as heart or breathing rate, or any changes in the tone, pitch or tremors in the voice.

The TUC has published a briefing note here (warning: word document).

See also:

Hat tip to Enrica Dente’s Lie Detection email list for picking the story up first.

Benefit cheats face lie detectors

liartrashOh please.

Here’s a great way to start your day: in the BBC headlines this morning (5 April), news that voice stress analysis will be used in job centres to test benefit claimants:

Lie detectors will be used to help root out benefit cheats, Work and Pensions Secretary John Hutton has said. So-called “voice-risk analysis software” will be used by council staff to help identify suspect claims. It can detect minute changes in a caller’s voice which give clues as to when they may be lying. The technology is already used by the insurance industry to combat fraud and will be trialled by Harrow Council, in north London, from May.

Said it before, and will no doubt say it again: voice stress analysis [pdf] is junk science. Relying on VSA in job centres will mean that genuine benefit claimants will be wrongly accused and fraudsters will continue to get away with it. So depressing to see the snake oil salespeople achieving success in the UK.

Photo credit: Patrick T Power, Creative Commons License

Voice Stress Analysis – ABC Primetime exposé

If you missed this last year, the ABC exposé of a company offering voice stress analysis technology is currently available on YouTube. As I wrote last April

The report featured a great interview with “Dr” Charles Humble, millionaire CEO of the National Institute for Truth Verification, the company that markets the tremendously successful Computer Voice Stress Analyzer. Humble claims that the system has a 98% accuracy rate in the field, but admits in interview that there are no independent scientific studies to validate this claim, is put under pressure on suggestions that the CVSA might have contributed to false confessions and is questioned about his title of ‘Dr’ which appears to add credibility to his claims to be marketing a scientific system. Humble eventually admits that his doctorate was granted by an Indiana Bible college after Humble took six hours of Bible classes.

Hat tip to the Anti-Polygraph Blog!

I’m unsubscribing from Skype today

According to the BBC (14 December):

Callers using internet phone system Skype who might be tempted to tell a few porkies should beware – the user on the other end may have a lie detector. Skype is to offer the KishKish Lie Detector, which is made by BATM, as an add-on for customers.

It analyses audio streams over a Skype call in real time and illustrates the stress levels of the other person. […] Paul Amery, director of Skype developer programme said: “This is a really excellent application, and the kind of thing we want to see more of.”

Gah! Firstly, I hate the idea of either testing my friends – or being tested by them – during our phone conversations. Secondly, the KishKish site explains that the add-on is based on Voice Stress Analysis, which has been showntime (pdf) and time again – to be useless for detecting lies:

Voice Stress Analysis (VSA) is a type of lie detector which measures stress in a person’s voice. The use of Voice Stress Analysis (VSA) as a lie detector became popular in the late 1970s and 80s. In the 90s the first Computerized VSA (CVSA) systems came to out to the market. The CVSAT is now the truth verification device of choice in the law enforcement community as the number of law enforcement agencies utilizing the CVSAT continues to grow dramatically, proving the viability of the system for twenty-first century crime detection. The CVSAT is also being utilized by the US Military in the global war on terrorism.

(a) VSA machines are NOT lie detectors, any more than polygraph machines are – they claim to measure indicators of stress, which is not the same as measuring deception.

(b) Who says the CVSAT is the ‘truth verification device of choice’ for law enforcement? It is not used in law enforcement agencies in the UK. VSA use is widespread in the USA, but just because many agencies use it does not mean that its ‘viability’ is proved. The US military are no longer using it, after a Pentagon study proved it detected lies at a rate of chance.

The bottom line: VSA is junk science. If you want a system with the same accuracy rate, I’ll sell you a penny – a coin toss gives the same success rate in detecting lies.

Paul Amery might want to see more of this kind of thing but hopefully he’s in a minority…

Proceedings of the Workshop on the Use of Autonomic and Somatic Measures for Security Evaluations

The most recent issue of the free online Journal of Credibility Assessment and Witness Psychology (vol. 7, No. 2, published June 2006) is a special issue, containing the Proceedings of the Workshop on the Use of Autonomic and Somatic Measures for Security Evaluations.  The entire issue can be downloaded as a (very large!) PDF file, or you can access the papers (well, summaries of powerpoints, rather than papers) individually via the JCAWP page.

Contents:

  • Polygraph Screening,  Donald J. Krapohl
  • Issues In The Study Of Polygraph Screening Techniques, Michael Bradley
  • Using the Polygraph in Employment and National Security David C. Raskin & Charles R. Honts
  • Emerging Technologies in Credibility Assessment, Andrew H. Ryan, Jr.
  • Toward a Neurocognitive Basis of Deception, Ray Johnson, Jr.
  • The Polygraph: One Machine, Two World Views, Stephen W. Porges
  • The Use of Voice in Security Evaluations, Harry Hollien & James Harnsberger
  • Voice Stress, James Meyerhoff
  • Evaluating Voice-Based Measures for Detecting Deception, Mitchell S. Sommers
  • Emerging Methods and Detecting Stress and Thermal Imaging, Dean Pollina
  • Body Odors as Biomarkers for Stress, Pamela Dalton
  • Radar Technology For Acquiring Biological Signals, Gene Greneker
  • The Physiology of Threat: Remote Assessment Using Laser Doppler Vibrometry John W. Rohrbaugh, Erik J. Sirevaag, John A. Stern, & Andrew H. Ryan, Jr.
  • The Gaze Control System and Detection of Deception, John A. Stern
  • Eye Movement-Based Assessment of Concealed Knowledge, Frank M. Marchak
  • Multimethod Assessment of Deception on Personnel Tests: Reading, Writing, and Response Time Measures, Andrea K Webb, Sean D. Kristjansson, Dahvyn Osher, Anne E. Cook, John C. Kircher, Douglas J. Hacker, & Dan J. Woltz

Tremors of the Trade – Investigative tool or troublesome black magic?

Another valiant attempt to persuade law enforcement not to buy into the Voice Stress Analysis snake oil: Warren J Sonne on Officer.Com (13 June) examines the utility of machines that claim to be able to detect lying through tremors in the voice, touching on three key issues:

1) The reasons why VSA is so apparently attractive to police departments:

I think that the first problem comes from the “Pac-Man” generation. Give us a high-tech toy and some practice, and within an hour we will defeat those lying criminals. […But] there are no quick fixes to criminal investigations, nor are there magic boxes to help detectives figure out who’s lying. There are no shortcuts to competent investigations — at least, there shouldn’t be.

2) The danger of false confessions:

There are many other cases that have been overturned by the courts involving people, some of them mentally challenged, who have confessed to crimes that they didn’t commit. As a result, many police departments across the country have instituted polices requiring the videotaping of all confessions, with some departments recording all, or nearly all, interviews. So, do we really need a Supreme Court decision here? Another Miranda? “You have the right to remain quiet and refuse to play the CVSA Game.”

3) The lack of any scientific evidence for the effectiveness of VSA:

Unfortunately, […] scientists and their studies have failed to demonstrate lie detection accuracy using CVSA at anything better than chance (flipping a coin).

Hat tip to Anti-Polygraph Blog for the link.

ABC Primetime report on false confessions and interview tactics

ABC Primetime last Thursday (30 Mar) screened a special report on false confessions and police interview tactics. The report featured a great interview with “Dr” Charles Humble, millionaire CEO of the National Institute for Truth Verification, the company that markets the tremendously successful Computer Voice Stress Analyzer. Humble claims that the system has a 98% accuracy rate in the field, but admits in interview that there are no independent scientific studies to validate this claim, is put under pressure on suggestions that the CVSA might have contributed to false confessions and is questioned about his title of ‘Dr’ which appears to add credibility to his claims to be marketing a scientific system. Humble eventually admits that his doctorate was granted by an Indiana Bible college after Humble took six hours of Bible classes.

You can view segments of the programme (including the Humble interview) here.

UPDATE: Not any more.  However, the story can still be accessed here.

Lying in wait for airborne criminals

Controversial device analyzes passengers’ voices
CNN, Thursday, January 5, 2006

Heightened fears of terrorist attacks have lead to a global beefing up of airport security, but controversial new measures are being developed that could see plane passengers screened by lie detectors. Features like iris and fingerprint scanning are now widely used by airlines, but with technology playing an increased role at frontiers, so-called E-passports containing biometric information about passengers are also becoming more commonplace. […] Now a new walk-through airport lie detector developed by Israeli scientists could throw up yet another layer of security to ensure potential hijackers or contraband smugglers do not gain access to international flights. The GK-1 voice analyzer, created by Israeli firm Nemesysco, requires passengers to don headphones at a console and answer “yes” or “no” into a microphone to questions about their travel plans.

This is a rehash of a story that first appeared on Reuters, November 17, 2005.

Arguments rage over voice-stress lie detector

The Arizona Republic Oct. 10, 2005

At least 20 Arizona law enforcement agencies are relying on a voice-measuring lie detector for criminal investigations even though experts say the device does not stand up to scientific scrutiny and may prompt innocent suspects to make false confessions. The Computer Voice Stress Analyzer, or CVSA, purportedly measures FM radio waves produced by muscles around the larynx. Deceptive answers cause stressful “micro-tremors” in the voice that are charted by the device’s software program, the manufacturer says. Yet, independent experts have consistently found the instrument to be dubious, at best, when it comes to separating truth from lies.

Psychophysiological and vocal measures in the detection of guilty knowledge

Matthias Gamer, Hans-Georg Rill, Gerhard Vossel, and Heinz Werner Godert
International Journal of Psychophysiology, article in press, available online July 2005

The Guilty Knowledge Test (GKT) and its variant, the Guilty Actions Test (GAT), are both psychophysiological questioning techniques aiming to detect guilty knowledge of suspects or witnesses in criminal and forensic cases. Using a GAT, this study examined the validity of various physiological and vocal measures for the identification of guilty and innocent participants in a mock crime paradigm. Electrodermal, respiratory, and cardiovascular measures successfully differentiated between the two groups. A logistic regression model based on these variables achieved hit rates of above 90%. In contrast to these results, the vocal measures provided by the computerized voice stress analysis system TrusterPro were shown to be invalid for the detection of guilty knowledge.

Area police agencies use voice “lie detector”

The Post-Dispatch, 05/14/2005

The urban legend goes like this: Police wire the gullible crook up to a copy machine, put a colander on his head and say the contraption is a lie detector. Intimidated, the crook confesses. Police say the latest lie-detecting gizmo – one that looks for vocal tremors in recorded conversations – is also eliciting confessions. But many scientists, including Washington University psychologist Mitchell Sommers, say the voice analysis devices are little more than $10,000 colanders. In distinguishing liars from truth-tellers, flipping a coin works better, according to a study Sommers performed.

The complete article is no longer available on the Post-Dispatch site, but can be read here (scroll down to 14 May 2005).