As a form of expert testimony, polygraph evidence is riddled with causation problems. A key concern is that it conflates correlation with causation; its design measures anxiety or arousal as manifested in blood pressure, galvanic skin response, and respiration, resting on the assumption that lying will provoke an anxious response. But some polygraph subjects are not anxious and do not exhibit physiological correlates of anxiety. And asking an innocent person “did you kill John Doe” may well evoke an anxious (but believed to be guilty) reaction. Additionally, measuring physiological responses to a question about murder against a control question like “did you ever steal anything as a child” may not be sufficiently discerning to determine serious lies from truth, since such a control question cannot pose the same level of stress that the real life questions can. Thus, the first problem is that the polygraph uses anxiety as a proxy for guilt, both overreaching and under-reaching, not recognizing that while there may often be some correlation between anxiety and guilt, the proof of actual causation is not as surefooted as claimed. The second problem is that it is incredibly difficult to create real-world consequences in control questions. The third problem is that countermeasures are potentially effective against the polygraph, competently disguising a “guilt” reaction
Moriarty, Jane Campbell
"Visions of Deception: Neuroimages and the Search for Truth,"
Akron Law Review: Vol. 42
, Article 4.
Available at: http://ideaexchange.uakron.edu/akronlawreview/vol42/iss3/4