Household surveys are increasingly moving from interviewer-administered modes to self-administered modes for data collection. With this move to self-administered questionnaires comes the problem of how to select a respondent within a household. Recent research has shown that many households do not appropriately follow within household selection procedures in mail surveys (Battaglia, et al., 2008). Yet this question has not been evaluated in web surveys of the general population or in mixed mode mail and web surveys. In this paper, we examine accuracy of within household selection using an oldest adult/youngest adult method in self-administered surveys. The frame for this study comes from a telephone survey conducted with Nebraska residents in which the oldest adult/youngest adult method is used to select the initial respondent. One year later, these telephone participants were randomly assigned to be followed up via a mail survey, a web survey, or one of two mixed mode surveys (mail with web follow-up, web with mail follow-up) using identical oldest adult/youngest adult household selection methods (n=566, AAPOR RR1 46%). Additionally, the sex of the previous respondent was provided to help facilitate identical selection of respondents. Yet surprisingly high rates of discrepancies occur when comparing the two sets of respondents. For example, 6.5% of respondents to the follow-up study differ on sex and 13.3% differ by at least three years of age. This paper will examine characteristics of people who appear to have followed the selection procedures appropriately compared to not following them; we also examine whether discrepancies in respondent selection vary across survey mode. We also examine predictors for discrepancies based on one criterion but not another (e.g., matching on sex but not on age). Preliminary analyses indicate that, as expected, mismatches are more likely to occur in larger households.