A case can be made that lie detector tests shouldn't be admitted in a criminal trial as evidence of guilt because they are too unreliable to establish proof beyond a reasonable doubt, but should be admissible when offered by a defendant since a defendant has no obligation to prove anything but does have a constitutional right to present exculpatory evidence. Courts generally do not accept that analysis, ruling that if the tests are unreliable when offered by the prosecution, they're also unreliable when offered by the defense. That view, which has the virtue of perceived consistency, is preferable to a rule that allows convictions to be based on the unsound science of polygraphy.

Probation agents love lie detectors. In some states they require probationers (particularly those on probation for sex offenses) to submit to regular polygraph examinations, and then require the probationer to pay for the tests. While probation agents don't have to prove anything beyond a reasonable doubt, it's a waste of resources (particularly those of the unfortunate probationer who has to pick up the polygrapher's tab) to require probationers to submit to a test that can't be shown to produce valid results.

Even more problematic are the so-called voice stress analyzers, a widely-used gadget that has zero scientific validity. That doesn't stop the police from bullying suspects into taking the bogus test, then telling them they've failed and insisting that they tell the "truth" if they don't want to spend a long, long time in jail. It's the sort of tactic that leads to false confessions.

The linked article suggests that courts are becoming more willing to accept the pseudo-science of polygraphy or voice stress analysis. Let's hope that's not a serious trend.