A local CBS television reporter went drinking to test various personal blood-alcohol devices. She found a wide disparity in readings among the different brands, showing I guess that you really shouldn't trust the things. What she fails to do, though, is ask why courts are then so reliant on them. She brought some patrol officers with her, and measured her results against the device she describes as "court-approved." But she never really questions whether or not that one is accurate (hint: it may not be).
She then says that the police officers who helped her with the story told her that "how a drinker scores in a field sobriety test is the real measure of inebriation."
In fact, this simply isn't true. The standard field sobriety test was adopted by NHTSA after one poorly administered test on 238 subjects in 1977. It's never been peer reviewed. One forensic expert in Georgia gave the test to 21 of his students, none of whom had a drop to drink. He then showed video of the tests to a group of police officers. They said they'd arrest nearly half of them.