observed sensitivity of the test and therefore biases upwards estimates of accuracy.
The other two studies that passed our screening (Raskin and Kircher, 1990; Honts and Amato, 1999) dealt with deception on preemployment screening tests. They both were pilot studies, had small sample sizes, allocated subjects to other treatment categories than just deceptive/innocent, and had a variety of other methodological problems. The results we could extract that pertained to accuracy were unimpressive, in the bottom 25 percent of the studies from which we extracted data.
One study deserves special attention because, although it did not meet our minimal screening criteria, it is the only available study that reports results from a real screening situation. Brownlie, Johnson, and Knill (1998) reported a study of 769 relevant-irrelevant polygraph tests of applicants for security positions at Atlanta International Airport between 1995 and 1997. The tests included four relevant questions, on past convictions for traffic violations or felonies, past bankruptcies, and use of marijuana during the past 30 days. As is typical with relevant-irrelevant testing, scoring was done by examiners’ impressions rather than any standardized method, a fact that makes generalization to other examiners very risky. The study reported results that correspond to an accuracy index of 0.81, a value well above chance, but still in the bottom 25 percent of the studies from which we extracted data.4
A desirable feature found in some screening studies is that examiners know neither which examinees are deceptive nor which of several questions a deceptive examinee will answer untruthfully (e.g., Barland, 1981; Correa and Adams, 1981; Honts and Amato, 1999; Raskin and Kircher, 1990; Timm, 1991). These studies mimic one aspect of true screening: the examiner is not certain which item is “relevant.” But in other respects these studies they are still a far cry from normal screening, in which the examinee has not been instructed specifically to lie, the list of possible deceptive answers is effectively infinite, and examinees may be deceptive about multiple items. In mock screening experiments, the mock transgression is highly salient, at least to all “programmed guilty” examinees, and everyone involved in the situation knows that the critical event is a specific staged transgression (even if they do not know the precise one). In typical real-life screening applications, there are a wide range of behaviors that might lead examinees either to admit minor infractions or to deny them, based on their individual perceptions of what the examiner “really” wants to know. Thus, examinees in actual polygraph screening may not know or may not agree about precisely what constitutes an honest and complete answer to some questions. In contrast, mock screening studies include a narrow range of issues that might be the target for deception, and subjects are assigned to deceptive or nondeceptive roles,