Using Personal Response Systems (PRS) in large lecture classes has been suggested as one way to encourage the active involvement of students in this challenging pedagogical environment. Positive correlations between PRS use and course grades have been shown, but few have attempted to determine what components of the active learning are most effective. Some factors that may contribute to the positive correlation include: increased time in which students are actively applying concepts in class, increased exposure to the type of questions asked by a particular instructor, and peer discussion opportunities that often accompany PRS use. We are most interested in the latter factor, as we believe that knowledge construction is social and peer discussion is an effective technique to promote conceptual understanding. By studying peer discussions, we can gain information on how to structure these social interactions so they lead to the most conceptual understanding. We have initiated a study of peer discussions with an experienced PRS instructor. This instructor has included several classroom scenarios where students respond to a PRS question, discuss the question with a peer, and then vote again. We are tracking several characteristics of the peer team (gender, comfort level in discussing with peer, familiarity with peer, comfort level with material) and we will determine the effectiveness of these discussions by 1) correlating them with the number of correct PRS responses and 2) listening to the recorded discussions of five groups.