NCJ Number
202093
Journal
Polygraph Volume: 32 Issue: 2 Dated: 2003 Pages: 107-126
Date Published
2003
Length
20 pages
Annotation
This article reports on a system for quantifying psychophysiological information obtained in comparison question polygraph tests (CQT's) that was developed on a sample of 181 confirmed deceptive and 150 nondeceptive field cases.
Abstract
The CQT uses a dynamic interview procedure in which stress responses to questions are observed. At least two classes of questions are present in a CQT exam. Relevant questions (questions related to the event at issue) may be answered with either deceptive or nondeceptive "yes" or "no" answers, and they are compared with a baseline of physiological responses established by presenting "comparison questions." The comparison questions are phrased in such a way that subjects feel they must be deceptive to complete the exam successfully. Comparison questions are intended to thereby elicit emotional arousal in both deceptive and nondeceptive suspects. The system proposed in this article uses permutation tests to obtain separate estimates of the likelihood of deception and nondeception for individual cases. These probabilities are then combined with base rate information to calculate an overall probability of guilt for each case. Assuming a 50-percent base rate of deception and conservative cutoff points for assigning deceptive (p=.90) and nondeceptive (p=.10) outcomes, 91 percent of deceptive cases (N=152) and 98 percent of nondeceptive cases (n=128) with conclusive outcomes were correct. The system was cross-validated on four groups of field cases. Conclusive outcomes were correct in 90 percent of deceptive cases verified by confession (n=97), in 92 percent of deceptive cases verified by urinalysis tests for drug infractions (n=49), and in 64 percent of deceptive cases independently verified by physical evidence or subsequent confession not associated with the polygraph test (n=11). Conclusive outcomes were correct in 88 percent of independently verified nondeceptive cases (n=16). The system was also cross-validated on a sample of laboratory cases drawn from three separate experiments. In the lab samples, 88 percent of conclusive outcomes were correct for both deceptive (n=33) and nondeceptive (n=25) groups. The proposed scoring system circumvents several logical problems with traditional approaches to quantification in CQT polygraphy. It also provides an effective means of obtaining accurate classification in approximately 90 percent of cases. 11 tables, 1 figure, and 33 references