In this paper we report on a case study of rumor transmission during a nationwide scandal via China's most popular microblogging service, weibo.com. Specifically, we explore dynamics of the rumor discourse by characterizing different statement types and their evolution over time. We examine the roles that different user groups play in the rumor discussions. Through qualitative and statistical analyses, our results identify seven reaction patterns to rumors and their different development trends. We reveal a three-stage pattern of the change of leadership during the rumor discussions. By connecting social theories on rumor transmission to the large scale social platform, this paper offers insight into understanding rumor development in social media, as well as utilizing microblogging data for effectively detecting, analyzing and controlling public rumors. Copyright 2013 ACM.