In this article, the author examines the character of the conversations generated in an elementary teacher group as they worked on mathematical problems together and analyzed their students' work. Two distinct forms of talk--"exploratory" and "expository"--were found. The first type of talk occurred most prominently when discussions centered on the teachers' own mathematical work and the second type when conversations centered on that of their students. By examining closely the few occasions when the groups' expository talk turned exploratory, the author explores how both the nature of the tasks and the range and type of facilitator conversational strategies can play significant roles in promoting and interrupting these conversational patterns to educational ends.