On-line discussion, while promising in theory, often turns out to be disappointing in application. Low levels of student participation are a particularly vexing and common problem. I argue that to overcome these chronic problems, instructors must recognize the collective action problems inherent in on-line discussion and address them, primarily through the use of selective incentives.
I test this contention using on-line experiments conducted on undergraduate political science students. The results of the experiments provide evidence of an underlying free-rider problem and support for the contention that selective incentives are an effective means of overcoming this problem. In another level of analysis, the experiment also shows that even basic student-to-student on-line discussion can be an effective teaching tool, rivaling the effectiveness of more costly traditional web sites.