Response error refers to the tendency to respond to items based on the perceived social desirability or undesirability of given responses. Response error can be particularly problematic when all or most of the items on a measure are extremely attractive or unattractive. The present paper proposes a method of (a) distinguishing among preferences even when all items are uniformly attractive, and (b) computing "intraindividual" reliability coefficients to detect and potentially delete respondents whose scores are insufficiently reliable. The method builds on the strategy of presenting items within item triads. (Contains 2 tables.)