Publication details
Hybrid Polygraph and Ocular-Motor Deception Tests for Screening and Specific-Incident Investigations
| Authors | |
|---|---|
| Year of publication | 2021 |
| Type | Chapter of a book |
| MU Faculty or unit | |
| Citation | |
| Description | We describe two experiments combining polygraph and ocular-motor methods to detect deception. The first evaluated a test covering four issues consisting of an automated polygraph and an ocular-motor deception format. 180 participants were randomly assigned to one of three conditions. One group stole $20 from a secretary's purse and lied about it. Another group stole the $20 and a ring from a desk and lied about both crimes. The third group was innocent answering all questions truthfully. Logistic regression combined features extracted to compute the probability of deception. The probability of deception was used to classify participants as guilty or innocent. On cross-validation, classifications were 92.2% and 90.0% correct for guilty and innocent participants, respectively. The second experiment evaluated a directed-lie protocol. 120 participants were randomly assigned to guilty (steal $20) or innocent conditions. All took an automated polygraph and ocular-motor version of the test. On cross-validation, decision accuracy was 87.1% for the innocent and 85.5% for the guilty. Both experiments assessed an indirect measure of blood pressure known as pulse transit time which was diagnostic, making significant contributions to the logistic regression models. Polygraph signals contributed significantly to the decision models and produced modest improvements in classification accuracy. |