bagwill@swe.ncsl.nist.gov (Bob Bagwill) (01/12/91)
A friend of mine couldn't get a clearance because he kept flunking the polygraph. Not because he lied, but because his results were too variable to classify. In science, it's common (bad) practice to throw out samples that are way off the curve, because they are "unrepresentative". As a result, the tester/researcher ends up selecting for testable-ness, rather than the trait or quality that they were supposed to be testing for. Sometimes the variance in the test is greater than the variance in the samples. Rather than throw out the test, it's easier to throw out the hard-to-test samples. It's like the man who looked for his wallet under the streetlight because they light was better there. I can imagine a company giving you a blood test, a personality inventory, a polygraph test, and a handwriting test, and then refusing to hire you if the results don't agree, on the grounds that you are unstable, are at least unpredictable. -- Bob Bagwill NIST Software Engineering Group/NCSL Technology Bldg, Room B266 bagwill@swe.ncsl.nist.gov Gaithersburg, MD 20899 voice (301)975-3282 fax (301)590-0932