DoDPI Banner for Research Papers

Abstract

CESTARO, V. L., A comparison between decision accuracy rates Obtained using the polygraph instrument and the Computer Voice Stress Analyzer (CVSA) in the absence of jeopardy. August 1995, Report No. DoDPI95-R-0002. Department of Defense Polygraph Institute, Ft. McClellan, AL 36205.

This two-experiment study was designed to (1) validate the underlying electronic theory of operation of the Computer Voice Stress Analyzer (CVSA), and (2) examine the decision accuracy and agreement rates using the traditional polygraph instrument and the CVSA. During experiment 1, the CVSA input/output was evaluated using simulation signals from laboratory test generators. During experiment two, forty- two subjects took psychophysiological detection of deception (PDD) examinations administered with the polygraph and CVSA instruments, within the context of Peak of Tension (POT) tests using numbers between 3 and 8. Half of the subjects were tested with the polygraph instrument, then the CVSA instrument. The remaining half were tested using the instruments in the opposite order. PDD and CVSA based POT tests were blind-evaluated by four independent examiners for each instrument. The frequencies of accurate determinations made using each instrument were compared using proportionality tests. The laboratory simulations established that the CVSA performs electrically according to the manufacturer's theory of operation. However, the CVSA instrument and associated processes were less accurate than the polygraph and PDD processes tested in similar circumstances (38.7% vs. 62.5%). Interrater reliability, assessed using a multiple rater Kappa test, showed agreement among all blind evaluators within each instrument category was significantly better than chance (p <.05). These data indicate there may be a systematic and predictable relationship between voice patterns and stress related to deception.

Key-words: voice stress, micro-tremor, CVSA, PSE, PDD, laryngeal, cricothyroid, cricoarytenoid, thyroarytenoid, vagal, interrater reliability

Director's Foreword

In recent years, there has been a renewed interest in voice stress applications in Psychophysiological Detection of Deception (PDD) testing. Due to the active promotion through advertising, selling, and training of examiners, the Computerized Voice Stress Analyzer (CVSA) has gained particular prominence among devices designed to use the voice in PDD. The implications for traditional PDD testing are obvious. If a device such as the CVSA is found to be valid, it would be heartily endorsed by Forensic Psychophysiologists. Technology of this nature could theoretically be used to supplement, or even supplant, the equipment and sensors currently used by PDD Examiners. Unfortunately there is a virtual dearth of scientific research regarding the CVSA; from either the promoters or the greater scientific community.

Thus, in accordance with the congressionally mandated requirement for DoDPI to study new technologies and sensors, this important study is the second in a series designed to address issues of voice stress in PDD testing. Specifically, this study was designed to (1) determine if the underlying electronic theory of operation of the CVSA is valid; and (2) compare the decision accuracy and agreement rates between the traditional polygraph instrument and the CVSA. What makes this study unique is that it represents the first comparison of decision accuracy between the CVSA and the traditional polygraph instrument. Such comparisons are scientifically very difficult (and hazardous) due to the normal differences in test formats, pretest interviews, and other procedural differences required by the two approaches. Consequently, a peak of tension format was employed to avoid these differences required by other formats.

As expected, this study found that in laboratory simulations, the CVSA does perform electrically according to the manufacturer's theory of operation. Although the decision accuracy utilizing the traditional polygraph and its procedures was significantly higher than that of the CVSA approach (62.5% vs 38.7%, respectively), both performed at a level greater than chance. While the validity of the traditional polygraph approach has long been established and this study certainly does not establish the validity of the CVSA, it does justify the necessity for further research into the CVSA. As resources permit, DoDPI intends to pursue further scientific inquiry into using aspects of the human voice in the detection of deception.

John R. Schwartz

Acting Director