Can a polygraph actually detect lies?

Not directly. A polygraph measures physical changes such as heart rate, blood pressure, breathing and sweating, and an examiner infers possible deception from them. The Journal of the American Academy of Psychiatry and the Law states that a specific physiological lie response has never been demonstrated and is unlikely to exist. Experts differ on how much the inference is worth, but most agree it is far from foolproof. [1] [2]

The most widely used method is the comparison-question technique, developed in 1947. It compares responses to questions about the investigation, neutral questions, and questions about past behavior designed to elicit lies. The theory is that a guilty person reacts more strongly to the relevant questions and an innocent person reacts more strongly to the control questions. A second method, the concealed-information technique, looks for arousal at recognizing a detail only a guilty person would know, such as a specific murder weapon. [1]

Because the signals reflect arousal, they can also appear when people are nervous but truthful. Dr Sophie van der Zee told the BBC that a polygraph is quite good at identifying lies but not very good at identifying truths, since the test itself is stressful. [1] [3]

Reported accuracy ranges from around chance to near 100 percent, depending on study design. A National Research Council review in 2003 is cited by both sides. The American Psychological Association summarizes it as finding that the comparison-question technique identified lies about 70 percent of the time, with an unknown false-positive rate, a weak scientific basis and low-quality research. The psychiatry journal reads the same review as putting accuracy at 81 to 91 percent, which it calls well above chance but well below perfection. [1] [2]

Critics include William Iacono and Gershon Ben-Shakhar, who reviewed later research in 2019. They concluded that research quality had changed little, that the 2003 conclusions still stand, and that polygraph supporters have misrepresented the report. Aldert Vrij argues the premise is flawed because there is no sound theory linking deception to the arousal measured. Proponents such as Don Grubin say that with a well-trained examiner, a properly conducted test and quality controls, accuracy is estimated at 80 to 90 percent, higher than the average person's ability to spot lies. An industry meta-analysis reported 89 percent, but critics note it was not peer-reviewed or independent. The APA says most psychologists and scientists see little basis for the validity of polygraph tests. [4] [3] [1]

Test type matters. Single-issue tests and concealed-information tests are described as more accurate, while multi-issue tests, common in screening, are less so. The National Research Council warned against using polygraphs in employment screening. Grubin also notes that testing victims is problematic, since recounting a traumatic experience would be expected to cause arousal anyway. [2] [5] [3]