Abstract Most surveys begin with questions designed to determine whether the sample household includes any members of the survey’s target population. However, the screening questions can miss eligible household members. The underreporting of eligible household members may reflect interviewer motivation, respondent motivation, or some combination of the two. We did an experiment to test several hypotheses about this phenomenon, which we call motivated underreporting. The experiment was done as part of a telephone survey, targeting 35- to 55-year-olds. The experiment varied whether the advance letter sent to sample households revealed the target age range for the study, whether the screening questions asked about specific age groups or collected information about each household member, and whether the interviewers received a bonus for each screener they completed, for each main interview, or no bonus. The main finding is that the format of the screening question affected the proportion of households screening into the survey as well as the response rates to both the screening and main interviews. The version of the screening questions that required respondents to roster the household members yielded the highest proportion of eligible households, but the lowest response rates. There is clearly a potential trade-off between coverage and nonresponse errors. These findings suggest that respondents are the source of underreporting of eligible household members. In addition, interviewers with higher screener completion rates tended to find fewer households with eligible members, suggesting that they may have screened out eligible households to achieve high response rates to the screener.