This study examined how characteristics of Collaborative Peer Video Analysis (CPVA) events are related to teachers' pedagogical outcomes. Data included 39 transcribed literacy video events, in which 14 in-service teachers engaged in discussions of their video clips. Emergent coding and Statistical Discourse Analysis were used to analyze the data. Findings suggest that particular types of clips (assessment vs. instruction), the purpose for sharing (problems vs. successes), and conversation-turn characteristics--including aspects of pedagogy discussed, aspects of disciplinary knowledge discussed, sources of knowledge, and conversation actions--predicted teachers' outcomes, such as generation of specific categories of pedagogical ideas and their consideration of applying these ideas in their future literacy pedagogy. Additionally, based on the explanatory model, most of the differences in whether teachers generated ideas or considered applying them in future occurred within a CPVA event (across conversation turns) rather than across events.