The polygraph exam typically has two main uses: for incident-specific questioning, like the police use in investigations; and for screenings, such as the ones employed by government agencies.
1) In criminal investigations in the U.S., the polygraph is used to determine if statements by suspects concerning a specific crime are truthful or deceptive. The polygraph test in this case is specific to the incident in question. There are two types of questions asked. First, examiners will ask “relevant” questions pertaining to the specific incident under investigation, such as, “Did you rob the Quik Mart on June 14, 2014?” Examiners also use “comparative” questions (called probable lie questions), which are used to establish a reaction to intentionally vague or difficult questions, or are questions that are impossible to answer truthfully with an unqualified negative answer. For example, one might ask, “Between the ages of 18 and 28, did you ever lie to someone in authority?”
At the conclusion of the test, the polygraph examiner compares the physiological reactions to the various question types. A subject who reacts more strongly to relevant questions is considered deceptive. The accuracy of this type of test is influenced by the skills of the examiner. Under optimal conditions, the accuracy of probable-lie tests in incident-specific criminal investigations is approximately 90 percent. (Source: American Polygraph Association Ad Hoc Committee on Polygraph Techniques, 2011)
2) Various U.S. government agencies use polygraph test to screen job applicants, employees, sex offenders and parolees. In contrast to incident-specific criminal investigations, relevant issues are more general, such as: “In the past 90 days, have you used any illegal drugs?” The generality of relevant questions in screening examinations is considered desirable because the test covers a wide range of illicit behaviors of concern to the agency. However, the generality of the relevant questions may introduce ambiguity in the mind of the examinee about their guilt (“I haven’t used illegal drugs in past 90 days, but I used them six months ago, and I know that was wrong”). The generality of relevant questions also increases their similarity to comparison questions, which are intentionally vague and broad in scope.
It is expected that reactions to comparison and relevant questions are more similar in magnitude and less diagnostic in screening tests than specific-incident tests, increasing the risk of false positive and false negative decision errors. Under these circumstances, the accuracy of polygraph falls in range of 65 to 85 percent.