In the new Scientific American, Charles Seife (author of Proofiness) takes apart the Pew poll which apparently showed that atheists knew more about religion than religious people. Seife notes that the results are a lot less clear if the standard technique for representing uncertainty is used (only 6% of the sample self-identified as "Atheists") - and if "Nothing in Particulars" (10%), many of whom don't believe in God, are classed with Atheists.
The Pew study revealed less about our faith in God than it did about our faith in polls - which, far too often, is blind. (SciAm, Dec 2010, p33)
As you may recall, your trusty religion poll reader noticed much that Seife finds wrong with the way the poll was reported (along with things it takes a religious studies scholar to notice). You read it first here!