Results from a series of experiments aimed at improving the quality of openend questions in mail, web and telephone surveys will be presented. The experiments were conducted annually between 2002 and 2007 using random samples of students (n’s approximately 200-350 for each of the experimental panels) at Washington State University, with the aim of testing multiple mechanisms for improving the number of words, number of themes, and extent of elaboration that occur in different modes for the same or similar questions. Results reveal that the quality of open-ended responses can be improved for each of the survey modes, but require somewhat different mechanisms in each mode in order to achieve that goal. The overarching goal for conducting these experiments is to find ways of achieving equivalent quality for open-ended questions across survey modes.