We have reviewed the principal lie detection techniques currently in use (Table 13.1). I should make it clear that numerous other lie test formats have been expounded in the literature, many with exotic names such as The Matte Quadri-Track Zone Comparison Technique, The DoDPI Bi-Spot and Tri-Spot Techniques, The General Question Test, The S-K-Y Test, and so on. In a recent text,1 a polygrapher named J. A. Matte lists a total of 18, but they are all minor variations on these same themes, involving in diverse combinations the same implausible assumptions.
We have seen that no wholly satisfactory assessment of the validity of any of these methods has as yet been done. Such an assessment would require some large police agency to test every possible suspect over a period of months or years, leaving the tests unscored so that they cannot influence the subsequent investigations, and omitting any posttest interrogation of the suspects by the polygraph examiners. Once any case has been cleared, yielding certain knowledge of which suspects had been truthful and which deceptive, knowledge uncontaminated by the polygraph results, the tests of those suspects would then be scored by polygraphers ignorant of the case facts. Because the existing field studies used polygraph-induced confessions as the criteria for ground truth, we know that the results of these studies provide only upper-limit estimates of lie detector validity. These upper-limit estimates suggest an accuracy of about 60% in detecting truthful responding and about 85% in detecting lying (Table 8.2) versus a chance value of 50%. It is certain that these estimates would be revised downward based on the results of an adequate field