WITH EVERY major spy scandal that hits the federal government, the federal reliance on polygraphs -- colloquially known as lie detector tests -- seems to grow. Polygraphs are now used not only in the context of specific criminal investigations, but also as a screening tool for employment in a number of federal agencies. Many employees of federal security agencies have to take polygraphs before being hired, and they are subject to testing during their employment as well. And this dependence on polygraph exams has developed despite serious questions as to the reliability of the "science" underlying them. Critics have long alleged that polygraphs rest on a kind of junk science and that they cast suspicion on the innocent, allow the guilty to escape detection, and create a false sense of security among agencies that cannot afford technologically driven complacency. Now a new report by the National Research Council, commissioned by the Energy Department, has added significant weight to these concerns -- concerns that ought to be tempering government's enthusiasm for this flawed instrument.

The report, which analyzes the controversy based largely on existing research, starts with the less-than-reassuring premise that the research to date has generally been of low quality. Polygraph testing, the panel contends, is intrinsically susceptible to error. It can be effective when used with respect to specific incidents -- as in criminal probes, where a suspect is asked whether he committed the crime. But when it is used for employment screening, "available evidence indicates that polygraph testing as currently used has extremely serious limitations . . . if the intent is both to identify security risks and protect valued employees." Actual spies may be able to employ "countermeasures" designed to beat the system. If polygraph testing is made so sensitive as to make avoiding detection more difficult, the result is that large numbers of loyal employees fall under suspicion as well. The report hypothesizes that in a population of 10,000 employees that includes 10 spies, 1,606 loyal employees will fail a test sensitive enough to identify eight of the bad guys. And if the test's sensitivity is set so as to reduce the number of loyal employees implicated to 40, it will then net only two of the spies.

One obvious conclusion from this sobering account is that more rigorous studies of polygraphs and their applications in government need to take place. A large number of people every year are put on administrative leave as a result of having failed -- or having not conclusively passed -- polygraph exams. And many people the government wishes to hire become off limits as well. The toll, in other words, is high -- both in human terms and in terms of the people who become unavailable to public service. To the extent this high cost is unnecessary, even counterproductive, people ought to know. And it is certainly wrong to go on expanding public use of the technology in the areas in which its effectiveness is most questionable. Those who defend the polygraph often cite the fact that people, believing that they will be caught by the lie detector, confess to significant breaches instead of facing the exam. But government's most sensitive employment decisions should not be made by placebo effect. Courts generally do not use polygraphs, out of concern about their reliability. Neither should the executive branch in areas where it cannot persuasively demonstrate the effectiveness of the tool.