IT WASN'T exactly surprising to learn that Dr. John F. Beary has reported that lie detector tests aren't a good way to determine whether someone is telling the truth: Dr. Beary is far from being the first eminent professional to make that judgment, and he won't be the last. His finding is noteworthy because it came in a report he made, as the acting assistant secretary of defense for health affairs, to Secretary of Defense Caspar Weinberger. Mr. Weinberger has before him two proposals for wider use of polygraph, or lie detector, testing, one initiated by the Pentagon itself and the other coming last month from President Reagan.

What's wrong with lie detector tests? What's wrong, as Dr. Beary explains, is that they measure stress rather than prevarication. So polygraph tests often misclassify truth-tellers as liars, and liars as truth-tellers. That's because some people show signs of stress even when they're telling the truth, while some cool-as-a-cucumber subjects can lie without skipping a heartbeat. That's why lie detector tests aren't admissible as evidence in court.

Despite all this, the president on March 11 ordered departments to draft regulations enabling them to require polygraph tests and to punish employees who refuse to submit to them. This goes far beyond even what the Pentagon has been considering since last fall; under its proposal, tests were still supposed to be voluntary, and adverse inferences were not supposed to be drawn against employees solely because of polygraph evidence. The argument is made that polygraph evidence is sometimes accurate. So is evidence obtained by voodoo, but that doesn't make it reliable enough for government to use.

Why is the Reagan administration ordering such widespread use of such an unreliable device? The aim, it is said, is to plug leaks. We'll stipulate that leaks are a serious problem, not only for an administration's political standing but occasionally for the national security as well. But we note that the government official in charge of these matters, director Steven Garfinkel of the Information Security Oversight Office, testified that only "half a dozen" leaks had been reported to his agency in the past three years, and that some of those did not result in stories in the press. And the head of the intergovernmental panel which recommended that the president order the measures that he did testified that "we have never suggested that it's a problem that has increased greatly in severity in recent years." So why this departure from longstanding practice? Why this reliance on an inherently faulty means of determining the truth?

Congress is looking into the matter, and perhaps the White House will have second thoughts. This is one issue on which we hope Mr. Weinberger and his colleagues will engage in a little constructive foot- dragging, in the hope that the White House comes to its senses.