Coverage of the general population in a household survey depends not only on the unit-level sampling design, but also on the method used to select an eligible respondent – the so-called “designated respondent” – within each household. In 2004, researchers from Westat (Rizzo, Brick, and Park) described a new method for within-unit respondent selection based on the number of eligibles in a household, which in theory appears to be a accurate and cost-effective approach to respondent selection. Our paper will report on our use of a hybrid version of the Westat method that has been implemneted in numerous Associated Press/GfK Roper dual frame RDD surveys (combined n > 10,000) during the past three years. The method generates a much better overall unweighted female/male selection of eligibles than other non-Kish selection methods. However, unexpected results consistently occur whereby there is a significant interaction between (a) gender of the interviewer and (b) gender of the respondents selected. The patterns observed were found to be associated with various types of misselection errors and a form of respondent-related measurement error.