This work-in-progress paper introduces the concept of Respondent Behavior Logging (RBL), consisting of static and dynamic models that conceptualize respondent behavior when filling in online questionnaires. It is argued that web-based survey design may benefit from logging as a technique for evaluation, since such data may prove useful during re-design of questionnaires. Although other aspects of online surveys have attracted considerable attention both in industry and in literature, how the Web may leverage new and innovative techniques to support survey design is still underexplored. Some preliminary results are reported in the paper, and issues are raised regarding how to appropriately evaluate and demonstrate the qualities of the RBL concept as a means for survey re-design.