We present a laboratory experiment to study the formation of dynamic coalitions in a bargaining setting where the current status quo policy is determined by the policy implemented in the previous period. Our main experimental treatment is the ability of subjects to negotiate with one another through unrestricted cheap-talk communication before a proposal comes to a vote. We compare committees with no communication, committees where communication is public and messages are observed by all committee members, and committees where communication is private and any committee member can send private messages to any other committee member. We find that the ability to communicate has a significant impact on outcomes and coalitions. When communication is public, committees more frequently agree on outcomes which give a significant fraction of the resources to every member. With private communication, we observe a significant increase in the share of allocations that give a positive amount to a minimal winning coalition. When either type of communication is allowed, dynamic coalitions emerge more frequently and majoritarian coalitions last longer. The content of communication is correlated with outcomes and with the persistence of a dynamic coalition. These findings suggest a coordination role for communication that varies with the mode of communication.