Developing effective public education sites for the World Wide Web requires an understanding of both learning theories and what appeals to leisure learners. Research indicates that active learning modes are most effective (Hein 1998). But at Museums & the Web 2001, IBM researchers presented findings that leisure learners prefer passive entertainment experiences over more demanding interactive experiences (Karat, et al. 2001). Educational web developers are left with the disturbing proposition that they can create effective learning materials, but that leisure learners will not consume them. This paper reports on a web site evaluation project that seeks to clarify the different factors involved in successfully engaging leisure learners. This study focuses on six educational Web activities, all developed by Educational Web Adventures (Eduweb) in collaboration with its clients. They represent a wide range of approaches to structuring Web-based educational activities and cover a broad spectrum of subject content, including art history, American history, and natural science. They include an encyclopedic/reference site, a simulation modeling a process, role-playing games, and creative play activities. All of the sites are aimed at a middle school or above audience. The study uses quantitative and qualitative data from server log analysis and exit surveys of Web visitors. Duration of visit from the log analysis is a rough measure of holding power. The exit survey solicits quantitative feedback - demographics, personal preferences in Web activities, and evaluation of the Web activity in question - as well as qualitative feedback about the user's experience. We expect to correlate these responses and user evaluations of each project and draw some tentative conclusions about the appeal and effectiveness of various formats to different audiences for active learning on the Web.