Can a machine ever really know that someone is lying? Since more than a century ago, the polygraph has been making that very claim its scribbled pens and electronic traces proclaiming to expose concealed truths from the body’s involuntary responses. Still, despite its persistent use by police and staple position within popular imagination, the scientific community remains sharply polarized about whether such squiggled lines indicate deception or only reflect the physiological aftershocks of anxiety.
The contemporary polygraph measures three main indicators of autonomic arousal: respiration, cardiovascular activity, and skin conductivity. Strain gauges on the chest and abdomen monitor breathing; a sphygmomanometer cuff or plethysmograph measures blood pressure and pulse; electrodes on the fingertips monitor changes in sweat gland activity, which change skin conductance. The underlying assumption is that deception stimulates the sympathetic nervous system increasing heart rate, changing breathing patterns, and increasing sweating. But as psychologist Leonard Saxe wrote, “the idea that we can detect a person’s veracity by monitoring psychophysiological changes is more myth than reality.” Honest people can have the same stress reactions as liars, and skilled deceivers can keep physiological calm.
Accuracy estimates range widely. A Department of Defense meta-analysis set polygraph sensitivity at 59 percent and specificity at 92 percent, whereas the National Academy of Sciences has stated accuracy between 55 to 99 percent varying by context, examiner proficiency, and questioning style. In laboratory conditions, some tests have reported scores in excess of 80 percent, but field conditions bring uncontrolled variables into play medical states such as autonomic neuropathies, medication such as beta blockers, or even minimal countermeasures such as controlled respiration can dampen or obscure responses. Such weaknesses are the reasons why the polygraph’s most popular format, the Control Question Test, has been accused of being dependent on fear or anxiety as surrogates for guilt.
Courts have long exhibited this suspicion. Since the 1923 Frye ruling, the majority of jurisdictions in the United States have excluded polygraph evidence unless stipulated to by both parties. The Supreme Court, in United States v. Scheffer, reaffirmed that “the scientific community remains extremely polarized about the reliability of polygraph techniques.” Even when admitted, judges may exclude results under Federal Rule of Evidence 403 if prejudice outweighs probative value. Prosecutors sometimes contend that credibility assessments are the jury’s province, and expert testimony regarding veracity threatens to usurp that function.
The restrictions of peripheral physiological measures have stimulated interest in brain-based techniques. Functional magnetic resonance imaging (fMRI) identifies deception through detection of changes in blood oxygenation in areas like the dorsolateral prefrontal cortex, which exhibit increased activity during deception. In experimental settings, fMRI has attained accuracy rates higher than 75 percent, with some research collecting 80–90 percent. Nonetheless, as federal courts in United States v. Semrau and Wilson v. Corestaff Services held, such findings are not “ecologically valid” actual error rates in the real world are unknowable, sample sizes are small, and experimental paradigms frequently ask subjects to lie, possibly measuring compliance with tasks instead of actual deception.
Electroencephalography (EEG) provides another option, recording event-related potentials such as the P300 wave, which occurs approximately 300 milliseconds following the identification of a significant stimulus. The Concealed Information Test exploits this concept to identify whether a subject is aware of crime-specific information. While EEG is cheaper and more transportable than fMRI, initial P300-based deception detection exhibited mixed accuracy rates, ranging from 40 to 80 percent, and is vulnerable to countermeasures. More advanced analyses, like the multifaceted electroencephalographic response (MERMER), seek to enhance robustness by taking the sum of several neural signals.
None of the new technologies satisfies the rigorous Daubert standards for admissibility, however, especially the requirement of a known error rate in target populations. As the Department of Justice has made clear, in the absence of large-scale, well-controlled clinical trials, both polygraph and neuroimaging-based techniques are still scientifically untested for forensic application. Moral issues complicate adoption further: brain scans challenge cognitive privacy, and the threat of “mind reading” in courtrooms troubles scientists and civil liberties activists.
For the time being, the polygraph survives as a detection device, in particular for federal employment and national security purposes, where agencies like the FBI continue to need it. Its survival is not only due to the appeal of technological reliability in lying detection but also due to the lack of an altogether confirmed substitute. Whether further advancements in neuroscience or artificial intelligence will be able to transcend these scientific and legal challenges is an open and vigorously debated issue.
