How Did the Chapman Study Demonstrate Voice Stress Analysis Validity for Law Enforcement?
In 2012, Professor Emeritus James Chapman conducted a retrospective analysis regarding the use of voice stress analysis (VSA) in law enforcement settings. As the Former Director of the Forensic Crime Laboratory at the State University of New York at Coming (SUNY – Coming), Chapman was interested in finding out how effectively VSA technology could detect signs of deception in the human voice during criminal investigations. He knew law enforcement had faced challenges when using previous technologies designed for deception detection, such as the polygraph. Not only were these techniques cumbersome and time-consuming, but their results were not entirely reliable and confession rates were rarely higher than 50 percent. However, VSA promised to overcome these flaws and provide a more reliable truth verification method.
But as Chapman observed more and more law enforcement officers relying on VSA to support their investigations—greatly increasing rates of valid confessions—he was dismayed to note that previous scientific studies had largely overlooked the benefits of this technology. Therefore, he decided to undertake a large-scale study to evaluate the performance of VSA in real world circumstances, over the course of 18 years. Specifically, Chapman set out to either confirm or invalidate three hypotheses:
- That the rates of stress and confession are interdependent.
- That VSA is a reliable tool in efforts to discern between instances of deception and no deception.
- That the confession rates of guilty individuals are affected by the severity of the possible consequences.
Award-winning neuroscientist Marigo Stathis worked with Chapman to review and confirm the scientific validity of the resulting paper, which they entitled “Field Evaluation of Effectiveness of VSA (Voice Stress Analysis) Technology in a US Criminal Justice Setting.” Their research served to verify the effectiveness of Computer Voice Stress Analysis (CVSA®) in criminal investigations, with a focus on the usefulness of this technology in obtaining admissible confessions from suspects.
The Wide-Ranging Study Population
Chapman’s study is particularly convincing because he designed a retrospective study that covered such a long period—18 years—and included a large number of wide-ranging cases and subjects. He began with over three thousand possible studies. This list was narrowed down to about 2,100 that met a basic set of inclusion criteria:
- The case was a criminal investigation—not a questioning in a non-criminal investigation, an employment screening, or a verification of witness testimony.
- A confession was a potential outcome of the questioning.
- The exam was conducted by a trained VSA expert using a structured set of questions.
Having narrowed down the case population, 236 cases were randomly chosen to be included in the study. The demographic characteristics of the interview subjects varied widely, with ages ranging from 5 years to 74 years. About 84 percent were male and 16 percent were female. They could be characterized as criminals, suspects, defendants, persons of interest, those serving court-ordered mandates (involved in child protective situations), and/or alleged victims. Some had a previous criminal history, while others did not. Some were well-educated and/or wealthy, while Chapman categorized others as “below normal intelligence” or “indigents.”
The criminal investigations included in the study also spanned a broad range of crime types. The most common crime was murder (18.2 percent), followed by rape (15.8 percent) and grand larceny (24.9 percent), but other crimes—like arson, assault, robbery, criminal mischief, and narcotics—were also included in the sample.
In 91 percent of the cases, the VSA examination was conducted when the legal authorities had reached an impasse in the investigation. Even though the VSA examinations occurred over an 18-year period, they all followed the same basic process. First, the VSA examiner was briefed by law enforcement officers. Then, the examiner conducted a pretest interview with the subject, followed by the structured set of questions for the actual VSA test. Sometimes, this was followed by a retest to ensure certain questions were truly associated with stress. Finally, each exam resulted in one of two outcomes: “no stress indicated,” in which the subject was cleared, or “stress indicated,” which led to a post-exam interview of the subject to determine the reason for stress (and potentially obtain a confession).
Study Results and Implications
Out of all the examinations, the VSA technology detected stress in 92 percent of the cases. In these cases, 96.4 percent of the subjects later made self-incriminating confessions—a much higher statistic than the 20-50% confession rate research has shown occurs in standard police interrogations. This provides solid scientific evidence to confirm Chapman’s second hypothesis, but just in case, a variety of statistical tests were conducted to rigorously verify the results. Based on these tests, he was able to scientifically validate that the probability of achieving these results was significantly greater than chance. Chapman’s statistical analyses also confirmed his second hypothesis that stress and confession are, indeed, interdependent variables.
In order to verify his second hypothesis—the accuracy of VSA technology—another statistical test was conducted, creating a model that would predict the expected proportion of this population that was deceptive. This number was 91.7 percent, which means there were no false positive results. Conversely, the model also demonstrated there had been no false negative results.
Finally, data from each case was analyzed to attempt to correlate the likelihood of confession with the crime consequences—his third hypothesis. He found that a confession was most likely when the consequence was of lower severity.
Overall, Chapman’s study broke ground in the field of truth verification by providing solid scientific evidence that VSA technology is a reliable tool for use in criminal investigations. Some previous research had provided a theoretical basis for the technology, but this retrospective analysis verified VSA’s usefulness in real-life law enforcement settings. The evidence is particularly strong because it covers 18 years of investigations on subjects who vary widely in age, gender, and personal history. Thus, law enforcement officers need only to look at the convincing results of Chapman’s long-term retrospective study to see the solid scientific support for VSA truth verification systems.