Here's one thing i've learned to help me distinguish a control from a relevant question... this is for a pre-employment poly with direct-lie questioning:
Ask yourself if this: would have a 'desirable' applicant done this? ie. stolen from an employer (most people have stolen minor things)... driven after drinking (any amount)... committed an undetected crime. These are all controls.
When the seriousness of the questions increase, ie. a serious crime is defined, stolen something in particular or a large amount of money or mercc, it becomes relevant.
Trout,
This is an excellent point. We have already written something very similar for the next edition. We plan to suggest that readers ask themselves "Would even the current employees of the agency for which you are taking the 'test' all be guilty of the misdeed in question?" Remember, control questions are "probable lie" questions. The polygrapher assumes that everyone, even the most desirable applicants, will not be able to answer "no" with complete certainty.