The BARB audience measurement system has the reputation of being the most sophisticated in the world. But because of its methodologies and its quantitative nature the advertising industry has, to a greater or lesser extent, been forced into making a number of assumptions: - the peoplemeter button is pressed so the viewer must be paying attention - someone is watching, they must want to do so - impacts have the same value irrespective of where they are placed while we may deny that we make these assumption in practice we all do. If we don't then why do we continue to put average OTS figures at the bottom of media plans. If we suspect that some programmes or some spots have a greater or lesser value, then surely the effective number of impacts would be adjusted to reflect variations. On the other hand, if these assumptions are accepted as broadly true, and that it doesn't matter what is bought, that a rating is a rating, think just for a second about the enormous blind faith which is being paid to BARB. A system which recently was reporting 16% of all viewing being accounted for by uninterrupted viewing sessions of eight hours or more! The point of this paper is not to knock BARB nor has it been written to stimulate a technical debate, but to challenge normally high levels of inquisitiveness when common sense and common practice don't gel.