On Wednesday the House approved, 333- 71, a measure that many of the 333 as well as most of the 71 knew cannot achieve its goal. The measure was a broad grant of authority to the Pentagon to administer so-called lie detector tests. The goal, of course, is to detect spies before they can do the kind of damage alleged to have been done in the Walker case. But Congress needs to think harder about the means.

It's accepted that lie detector tests have some role in keeping secrets. The CIA and NSA routinely administer them to their employees, and to an extent that cannot be known the tests or the threat of them may well have prevented some spying activity. Polygraph testing, its proponents admit, is less than 100 percent reliable. Lie detectors detect not lies but stress. Skillful liars -- and presumably some spies -- can fool the machine. With good reason, the results of these tests are not admissible as evidence in court.

So the weapon the House proposes to rely on is faulty; moreover, its aim is imprecise. What is needed is not to authorize new fallible methods to test 4 million Pentagon and defense-contractor employees, but to zero in on the far smaller number with access to important secrets. Powerful institutional incentives exist to classify far too much material as secret and to clear far too many people to see it. This has made protection of genuine secrets more difficult.

One suggestion is to use a method applied by the Office of Management and Budget in domestic government: put hard-and-fast limits on the number of employees to be granted security clearances and on the amount of material to be classified. This would cause a lot of grousing and might lead to some wrong decisions. But under the pres man such as John Walker Jr., with access to geuine secrets, evidently escaped all scrutiny after he was first granted his clearance until he was implicated by family members. Authorizing the already overworked Pentagon security apparatus to give lie detector tests to 4 million people blurs any focus on the much smaller number of potential spies.