NCCA Banner for Research Papers


CESTARO, V. L. A comparison of accuracy rates between detection of deception examinations using the polygraph and the computer voice stress analyzer in a mock crime scenario. November 1996, Report No. DoDPI95-R-0004. Department of Defense Polygraph Institute, Ft. McClellan, AL 36205.

The accuracy and consensus of decisions rendered between examinations administered using the traditional polygraph instrument and the Computer Voice Stress Analyzer (CVSA) were examined. One hundred twenty subjects were given detection of deception examinations either on the polygraph instrument or the CVSA. Subjects were divided into two groups, with one group participating in a mock theft and instructed to be deceptive about their participation in that crime. The remaining group was told that a crime had been committed, but they did not participate in any way. This group was instructed to be truthful about not being involved in the theft. Half of each group of subjects were tested in a Psychophysiological Detection of Deception (PDD) examination using the polygraph instrument and the remaining half were tested using the CVSA instrument. Groups were counterbalanced for gender. Six blind scorers, three trained in polygraph use and three trained in use of the CVSA, independently rendered decisions for subjects' examinations. Neither group of blind scorers achieved overall accuracy rates better than chance levels. A significant interrater agreement among the three blind examiners within each instrument type was found using the kappa statistic for multiple raters (p <.05).

Key-words: voice stress, CVSA, PDD, DLCT, directed lie control, interrater agreement, jeopardy

Director's Foreword

Over the last twenty years it has been proposed, several times, that changes in human voice characteristics are indicative of deception. Several manufacturers have, over this period, marketed instruments purported to detect deception through analysis of human voice responses. Claims that the human voice response can be used to detect deception have, in the past, not been strongly supported in the research literature.

A relatively new device, the Computerized Voice Stress Analyzer (CVSA), has been marketed by the National Institute for Truth Verification. The CVSA has become increasingly popular, possibly because it is relatively easy to operate and operator training requires a minimal 40 hours of course work. Proponents of the instrument claim relatively high deception detection accuracy rates when the CVSA is used as taught by the National Institute for Truth Verification. These claims are, however, based primarily on anecdotal evidence rather than evidence obtained through rigorous systematic study.

This report describes one of the more scientific studies conducted using the CVSA instrument to date. The study was designed to compare the validity of data collected using a traditional polygraph instrument to that collected using the CVSA. It is the second in a series of studies being conducted at the Department of Defense Polygraph Institute. Those reviewing this study should remember that the results of one or two studies are rarely considered to be conclusive.

Michael H. Capps