NCJ Number
209869
Journal
Polygraph Volume: 34 Issue: 1 Dated: 2005 Pages: 47-61
Date Published
2005
Length
15 pages
Annotation
This study compared the traditional and two experimental methods of analyzing data from Relevant/Irrelevant (RI) screening tests for their accuracy and to the decisions from the original examiner.
Abstract
RI polygraph screening tests are widely used to help in police candidate selection processes. The RI screening test is a multiple-issue test involving three to five relevant questions and a series of buffering irrelevant questions. The current exploratory study compared three methods of analyzing data from RI screening tests to determine how each of the analysis methods performed against chance and to provide an indication of effect sizes. Stratified random sampling from a large archive produced a sample of 100 confirmed field RI cases. Supplemental data for the establishment of ground truth were obtained from urinalysis testing, police reports, official State records, and posttest admissions. Results indicated that the original examiner had an average by-question accuracy of 81 percent. In contrast, a blind scorer using global evaluation and an experimental method each produced 73 percent accuracy for the same questions. Intra-scorer reliability for the blind scorer’s decisions was statistically significant while reliability for the other decisions was not. The findings suggest that the RI screening process should perform better as a first stage of a successive screening process than as a stand-alone methodology for veracity. Additionally, the findings point toward blind scoring as the method of choice for independent quality reviews of RI screening data. Table, figures, references