One of the most promising features of web surveys is the possibility of embedding certain interviewer – like behaviors in self-administered questionnaires. We have been investigating one such behavior – giving feedback to respondents whose answers were unreasonably fast (“You seem to have responded very quickly. Please be sure you have given the question sufficient thought to provide an accurate answer.”). We conducted a web survey experiment in which 2463 opt-in respondents answered six questions about quantities (e.g., “Overall, how many overnight trips have you taken in the PAST 2 YEARS?”). Participants were randomly assigned to one of three conditions defined by whether and when they received feedback: the first time they gave a fast response, every time, or never. This feedback slowed response times (and presumably improved the quality of answers) for respondents that could receive the feedback relative to respondents that could not. However these benefits were evident for only certain respondents. After one or two prompts, some participants slowed down; but participants who were prompted three or more times never slowed down. The positive impact of the feedback transferred to another type of respondent short-cut: in three grid questions (options as columns, statements as rows), respondents prompted once or twice on the earlier questions were less likely to “straight-line” (i.e., select answers mostly in the same column) than participants prompted more often. Work currently underway tests the duration of this transfer effect. In general, the interactive character of the web may allow designers to increase data quality while reducing cost.