PolyCop,
I was referring above to CQT polygraphy in all its forms, whether in a screeing context or in interrogations regarding specific incidents. If some guilty/deceptive subjects can be successfully tricked into confessing through the use of CQT polygraphy, then certainly, it becomes possible to correctly classify subjects as truthful vs. deceptive at rates better than chance. But the underlying procedure still has no scientific basis whatsoever.
You are right that polygraph "tests" are not in all respects like a roll of the dice. The key difference is that in the latter, the odds of any particular outcome are demonstrably knowable. By contrast, in CQT polygraphy, absent a corroborable confession from the subject, the polygrapher cannot demonstrate with any knowable degree of confidence whether a particular person has or has not spoken the truth with regard to any particular relevant question.
I think it is appropriate to repeat here remarks I posted in the message thread,
NAS Polygraph Report:
What the NAS Report Says About the Accuracy of Specific-Incident Polygraph Testing The following is an excerpt from the conclusions of the NAS polygraph report (
p. 168 of the HTML version):
Quote:[font=Times,Palatino]Estimate of Accuracy Notwithstanding the limitations of the quality of the empirical research and the limited ability to generalize to real-world settings, we conclude that in populations of examinees such as those represented in the polygraph research literature, untrained in countermeasures, specific-incident polygraph tests for event-specific investigations can discriminate lying from truth telling at rates well above chance, though well below perfection. Accuracy may be highly variable across situations. The evidence does not allow any precise quantitative estimate of polygraph accuracy or provide confidence that accuracy is stable across personality types, sociodemographic groups, psychological and medical conditions, examiner and examinee expectancies, or ways of administering the test and selecting questions. In particular, the evidence does not provide confidence that polygraph accuracy is robust against potential countermeasures. There is essentially no evidence on the incremental validity of polygraph testing, that is, its ability to add predictive value to that which can be achieved by other methods.[/font]
Note that:
1) This estimate of accuracy does not specify what kind of polygraph tests, e.g., CQT vs. R/I vs. GKT "can discriminate lying from truth telling at rates well above chance."
2) The authors' conclusion that polygraph tests "can discriminate lying from truth telling at rates well above chance" is conditioned upon the subject population being similar to "those represented in the polygraph research literature," that is,
ignorant of polygraph procedure and countermeasures. Such ignorance cannot be safely assumed, especially with information on both polygraph procedure and countermeasures readily available via the Internet. 3) If the authors' conclusion that "the evidence does not allow any precise quantitative estimate of polygraph accuracy..." is correct, then it (a fortiori) follows that
software algorithms peddled by polygraph manufacturers such as Axciton and Stoelting that purport to determine with mathematical precision the probability that a particular individual is lying or telling the truth are worthless. 4) The authors conclude that "the evidence does not provide confidence that polygraph accuracy is robust against potential countermeasures." It is not safe to assume that anyone passing a polygraph "test" has told the truth.
5) The last sentence of the above-cited paragraph is the key one with regard to polygraph validity (as opposed to accuracy): "There is essentially no evidence on the incremental validity of polygraph testing, that is, its ability to add predictive value to that which can be achieved by other methods."
What this means is that there is no evidence that polygraph "testing" provides greater predictive value than, say, interrogating a subject without the use of a polygraph, or with a colandar-wired-to-a-photocopier that is represented to the subject as being a lie detector. The NAS's conlusion that "specific-incident polygraph tests for event-specific investigations can discriminate lying from truth telling at rates well above chance, though well below perfection" with naive subject populations is hardly a vindication for the validity of CQT polygraphy, and
those in the polygraph community are formally cautioned against publicly misrepresenting it as such, as you can expect to be publicly called out on it.