The April 3, 2008 Wall Street Journal published a letter by an economist re: the unreliability of polygraph testing that I thought worth of sharing:
Your article makes it crystal clear why a person should never consent to polygraph tests, which suffer from two types of errors: showing an innocent person to be guilty and showing a guilty person to be innocent. The article illustrates that a person is likely to fail the test because of perceived or actual guilt on subjects that aren't even being tested. The test i much more likely to show an innocent person to be guilty than a guilty person innocent. This is especially true if the person being tested is reasonably honest and worries about minor peccadillos, such as taking home a paper clip or coming to work a few minutes late. Because of this, polygraph tests encourage law enforcers who "know you are guilty" to keep digging until some question of innocence can be found on some (perhaps unrelated) activity.
Professor of Economics, Retired
Missouri Southern State University
See related article at Mother Jones: