Technical glitch raises more questions about polygraphs

Morrison Bonpasse has encouraged discussion on this blog of his study, “Polygraphs and Exonerations — A Promising Relationship,” on which he made a presentation at last month’s Innocence Network Conference. Bonpasse and I had several exchanges as he finalized his study, and he made some corrections and adjustments as a result.

After his presentation, Bonpasse said in an email that “the facts in my article speak for themselves” about the value of the polygraph in innocence investigations. But do they?

One key fact Bonpasse uses in his paper, which is available here, says that ”the 2003 National Research Council report, The Polygraph and Lie Detection, found an 86% accuracy rate for polygraphs on single issue testing.” But in an analysis of the polygraph’s reliability by a U.S. District Court judge in Atlanta in the case U.S. v. Ricardo C. Williams, the judge took issue with that 86% accuracy claim. It noted that the Research Council went on to say that the quality of the polygraph studies it reviewed “falls far short of what is desirable” and that the accuracy rates that resulted are “highly likely to overestimate real-world polygraph accuracy.”

Part of the problem with many polygraph studies is that they are conducted by people who directly or indirectly are on the payroll of the polygraph industry, whose first interest is profit, not truth.

McClatchy Newspapers Washington bureau reporter Marisa Taylor provided a good example of that in May 20 article here. The article reported that “police departments and federal agencies across the country are using a type of polygraph despite evidence of a technical problem that could label truthful people as liars or the guilty as innocent” because they haven’t been notified of the issue.

Taylor said the technical glitch in question produced errors in the computerized measurements of sweat in one of the most popular polygraphs, the Lafayette Instrument Co’s 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,” Taylor wrote.

The story noted that polygraphs, unlike medical or other computerized equipment, aren’t required to meet any independent testing standards to verify the accuracy of their measurements.

Although the LX4000′s problem has long been known, the article said, the experts or decision makers who should have been spreading the word or acting on it didn’t. One reason for that, Taylor reported in a separate story , might be that those experts — including full-time law-enforcement officers — are being paid by the machine’s manufacturer as consultants or dealers.

This can lead to serious conflicts of interest. Consider the two experts who developed the American Polygraph Association’s highly critical response to McClatchy’s findings about Lafayette’s LX4000, which accused McClatchy of exaggerating the problem and working for a competitor. McClatchy said both experts are on Lafayette’s payroll. While Lafayette’s competitors have used the LX4000′s problems to their advantage, they have done it quietly, lest someone start taking a closer look at potential flaws in their own instruments or raise more questions about the polygraph in general.

While Bonpasse’s study is interesting and he makes some good recommendations on how to make polygraph testing better, the polygraph still doesn’t pass scientific muster. For that reason, courts are not likely to accept a polygraph exam’s validity, which is what happened in the Ricardo Williams case mentioned above. In fact, in the two cases Bonpasse mentions that he has used the polygraph in an attempt to prove inmates’ innocence, both men remain in prison. So the polygraph is likely to remain a secondary investigative tool at best.

5 responses to “Technical glitch raises more questions about polygraphs

  1. I thank Martin Yant for his comments and the posting of the link to my article and the recent news about a claimed glitch in the measurements used in the Lafayette LX4000 polygraph system. Please be assured that I am not affiliated in any way with any of the manufacturers of polygraph equipment. Also, please note that although those manufacturers do aim for profits, just as police, prosecutors and defense attorneys aim for salaries, the competitive goal of those manufacturers is to produce a quality machine. The instrument which detects deception or non-deception the most accurately will capture the sales. Incidentally, the motto of the American Polygraph Association is “Dedicated to Truth.”
    I could have said in my email to Martin that the CASES also speak for themselves, as indeed they do. In the 213 exoneration cases researched for the article, the polygraph results pointed toward the truth of exoneration about 80% of the time. While 100% would be “desirable” to use the term from the Ricardo Williams case Martin cited, 80% is still surely useful. Doesn’t 80% compare well to the accuracy of eye witnesses? Thanks to the laws of probability, as the article explains, the accuracy rate improves when more than one exam is taken by the same person.
    My article did not argue directly for admissibility of polygraph results as evidence of truth in criminal trials, as is now the practice in New Mexico, and somewhat in other states with the use of stipulations. Instead, the article presents the results of real cases of real exonerees which support the increased used of video-recorded polygraphs, including Quality Reviews, by all parties in criminal cases. When the results are disputed by either side, the article recommends conducting a second polygraph. As the article notes, while DNA and/or fingerprints are collected in a small percentage of cases, there is at least one person who is a candidate for a polygraph in EVERY criminal case.
    Readers of this Wrongful Convictions blog are invited to send me additional information about any of the 213 exoneration cases, or information about other exoneration cases where polygraphs were used.

  2. Polygraph validity in general aside – here is an illuminating article about the obfuscation of Lafayette Instruments regarding problems with their LX4000 and LX5000 systems. http://spectrum.ieee.org/riskfactor/computing/it/it-hiccups-of-the-week-lie-detector-lies

  3. First of all, Morrison Bonpasse is wrong to refer to a “claimed” glitch in the LX4000 polygraph system. The manufacturer admitted that the glitch exists in a technical notice, albeit only after it found out that McClatchy Newspapers was going to publish a story about it.

    Second, I asked Mr. Bonpasse a long time ago to give me one example in which a polygraph exam was the determining factor in an exoneration case. I’m still waiting for his answer. It certainly doesn’t look like it will be either of the inmates he has had take polygraph exams. In one of those cases, the polygraph exam may actually have hurt his client’s innocence claim. The New Hampshire State Police issued a highly critical analysis of the exam. In addition to numerous deficiencies in test procedures that the state police believed made the test invalid, the report also indicated that the inmate may have utilized ”countermeasures” to make deceptive answers appear to be truthful. This is a standard law-enforcement accusation when a defendant takes a polygraph exam, and it’s hard to dispute it.

    Polygraph exams usually are a waste of time and money. If the inmate passes, prosecutors will criticize the way it was conducted. If the inmate fails, prosecutors will say it’s more proof of the inmate’s guilt.

    Giving an inmate a polygraph exam and declaring him innocent if he passes is a fruitless exercise, as Mr. Bonpasse’s client who passed an exam three years ago has discovered. It’s far better to concentrate on facts rather than speculation.

  4. Agreed. The word “claimed” was not the right word. I thought it was useful to use some adjective to reflect the uncertainty surrounding the glitch, but there’s no need. A glitch it is, and it should be fixed.
    Respectfully, it’s not for me to say where or whether a polygraph result was the determining factor in an exoneration. My article presents the record of accuracy for polygraphs when they were used for defendants and witnesses in exoneration cases, about 80%, and the article recommends that police, prosecutors and attorneys use polygraphs more often and pay more attention to the results.
    For the cases where pre-trial polygraphs pointed toward innocence, they were were not even close to determinative, as they were ignored, and such avoidance led to the wrongful convictions. I don’t know how much attention was paid to those same polygraph results after the trials during the struggles for exoneration. For the cases with post-trial polygraphs, there were surely several factors in each case which led to exonerations. Let’s not forget, too, that in 20% of the cases, the polygraph results were wrong, so they obviously were not the determining factor in any exoneration. To the extent they were noted, they helped thwart exonerations. As my article noted, there was no way I could control the quality of the polygraph exams which were conducted in the 213 cases. It’s my estimate that the 20% figure will be reduced over time to the extent that polygraph exams are video-recorded and subject to Quality Reviews and examinees are given opportunities to take a second exam. Like most forensic tools, they will never be 100% determinative of truth, but they are still very useful.
    Regarding my New Hampshire client whose first polygraph was inconclusive or invalid, the answer is just as my article recommends: The State of New Hampshire should allow him to take another polygraph exam, at no cost to the taxpayers, as he has requested.

  5. As a polygraph examiner, I think the argument if bigger than that to which you have limited it. How often polygraph has been useful after conviction, I don’t know. It’s difficult to quantify. How many advocates base their willingness to represent a client on polygraph results (and set the wheels in motion)? However, I can think of a number of cases that have resulted in defendants having their charges dropped, which means those incarcerated while awaiting trial were freed. Those cases will never be counted if you only look at those wrongfully convicted. (The goal is to avoid those from the start.)

    It is interesting that Morrison Bonpasse’s data supports what the NAS reported in regard to polygraph accuracy. That is, there is no (statistical) difference between their estimates.

    While Lafayette can defend itself, it is wrong to call a filter a “glitch.” A filter removes something from the data – by design. You essentially have two pieces of information in polygraph data: signal (helpful data) and noise (unhelpful data). The goal of a filter is to remove the noise. Thus, when data is filtered, we expect measurements to change on occasion. (And sure, there will be times when the filter removes part of the signal.) The reporter interviewed me, and I asked if there were any known cases in which this supposed glitch resulted in a truthful person being called a liar or a liar called truthful. There are no known cases. The only information she had was that the difference between the scores of filtered or unfiltered data was, at times, sufficient to change a decision to a non-decision. That is, a decision of truthful may have become inconclusive. The question is whether the filter helps examiners to better classify cases (correctly). The scientists at the University of Utah have found that their method of filtering data improves decision accuracy, so their “glitch” (in that measurements change when filtered) is beneficial.

    By the way, what is the accuracy of juries or judges? Assessing accuracy in those contexts suffers from the same problem as polygraph (seldom knowing ground truth). From my review of the literature, there is even less information on point than on polygraph, so the NAS criticisms would likely hold true there as well. There are only a couple of studies on accuracy, and they estimate judge and jury accuracy at about 87% – the same as polygraph. Again, those studies share the same generalization issues as polygraph, which the NAS criticized. Polygraph is just one piece of evidence, and if used correctly, as Morrison suggests, it can – and has proven to be – useful. Like any other test, it only provides information. What one does with that information may vary dependent on its purpose.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s