This project contributes to our understanding of the way economic agents communicate and coordinate in bargaining. It also develops a new bargaining model which can explain both the duration of negotiations that end without strikes and the frequency and duration of strikes. The research builds on the impressive past contributions of the investigator. The first part of the project is a theoretical investigation of how agents learn how to cooperate in repeated coordination games with complete information. These games allow a clearer characterization of what rationality means in the presence of multiple equilibria, and present the problem of learning how to cooperate in its purest form. The project studies how agents' optimal strategies depend on whether they begin with a common language in which to describe their actions and/or their positions in the game to each other. This research helps explain how social conventions evolve. It clarifies the concept of a rational response to multiple-equilibrium problems. The second part uses dynamic bargaining models to explain the timing and duration of strikes. Such models typically posit private information for one bargainer, and rely on differences in the costs of delaying agreement among his types to sort them in equilibrium. The project develops a bargaining model in which production usually proceeds during contract negotiations, and calling a strike requires an overt action by the union, with consequences vastly different from failing to accept the firm's last offer. Striking increases delay costs for both bargainers, and may also entail sunk costs.