Although video cases and video clubs have become popular forms of teacher professional development, there have been few systematic investigations of designs for such programs. Programs may vary according to (a) whether teachers watch videos of their own/their peers' instruction, or whether teachers watch stock video of unknown teachers; and (b) whether discussions are led by trained facilitators or by participants themselves. Using a factorial design, we defined four treatment conditions based on these possibilities, then assigned three groups of teachers to each condition. Teachers watched, scored, and discussed mathematics instruction according to each treatment condition's protocol. Evidence from groups' conversations and teachers' video analyses and lesson reflections suggest that the teacher-led, own-video condition is slightly superior to the other conditions.