In December 1923, a federal appeals court in Washington upheld a murder conviction by refusing to admit the results of what the court called the systolic blood pressure deception test: an early ancestor of the polygraph, on the theory that lying raises blood pressure in a pattern that truth-telling does not. The ruling turned on a narrower question than whether that idea was right: whether anyone had shown it to be. The test, the opinion found, had not gained general acceptance among the scientists who would know. For seventy years, that standard governed whether federal courts let in any new kind of scientific evidence. A different ruling replaced it in 1993. Five years later, the same underlying question, how reliable the polygraph actually is, reached the Supreme Court again.
This time it arrived by way of an Air Force base in California. Airman Edward Scheffer had volunteered as an informant on drug investigations, and in return submitted to periodic drug tests and polygraphs. In early April 1992, shortly after giving a urine sample and before the results came back, an examiner wired him to a polygraph and judged, from his breathing and blood pressure, that he was telling the truth about not using drugs since enlisting. The lab disagreed: his urine held methamphetamine.
Scheffer went absent without leave for thirteen days and was arrested by an Iowa state trooper on a routine stop. Court-martialed for the drug use, the absence, and a string of 17 bad checks, he wanted the polygraph's finding of no deception admitted, to back his claim that he had taken it unknowingly. Military Rule of Evidence 707 would not allow it: it kept polygraph results out of every court-martial, whichever side offered them. When the case reached the Supreme Court in 1998, the government defended Rule 707 on several grounds, but the central one was reliability: polygraph evidence had never been shown reliable enough to trust. The Court's own survey of the research agreed: some laboratory studies put the accuracy of the standard polygraph technique as high as 87 percent, while other studies put it at little better than a coin toss, 50 percent. Nobody could say which figure was closer to the truth. The military leaned on the test anyway: the Department of Defense alone ran it more than 400,000 times between 1981 and 1997.
Writing for the Court, Justice Clarence Thomas did not decide whether Scheffer had lied; he wrote that nobody could show the polygraph would have told them either way. No consensus had ever settled whether polygraph evidence was reliable, he found, and the scientific community remained as divided as it had been in 1923. The ruling left the device itself alone; investigators still use it every day. What it refused was narrower: treating an unproven verdict as proof, however badly either side wanted to believe it.
Seventy-five years and one change of legal standard separated the Washington courtroom from the Supreme Court's ruling on Scheffer's case. In between, the rule for admitting new scientific evidence changed completely. What never changed was the gap between the competing numbers: across seventy-five years of lawyers and judges, nobody had settled how often the device is actually right. A test that cannot grade itself has no business grading anyone else.