This study presents a new method for developing faking detection scales based on idiosyncratic item-response patterns. Two scoring schemes based on this approach strongly differentiated between scores obtained under honest vs directed faking conditions in cross-validation samples (rpb=.45 and.67). This approach is shown to successfully classify between 20% and 37% of faked personality measures with only a 1% false positive rate in a sample comprised of 56% honest responses. Of equal importance, this method does not result in a scale that meaningfully correlates with personality or cognitive ability tests. This study raises many questions about both the source and generalizabiltiy of the effect. Key directions for future research and improved scale development that may limit or enhance the utility of the idiosyncratic item-response method are discussed.