Studies of contextual effects on political behavior are plagued by concerns about internal validity. Perhaps of greatest concern are possible selection mechanisms that appear to present statistical support for contextual influence when social communication has no real effect. This paper presents an experimental framework for testing contextual effects that ameliorates these concerns through exogenous assignment to communication networks. This experiment allows for an analysis of the factors that make discussion partners influential. These factors can be divided into two categories: (1) characteristics of the dyad and its members; (2) characteristics of the residual discussion network. The most robust findings suggest that factors in this latter category play the greatest role in the likelihood that a discussion partner is influential.