We study costless pre-play communication of intentions among inexperienced players. Using the level-k model of strategic thinking to describe players' beliefs, we fully characterize the effects of pre-play communication in symmetric 2 2 games. One-way communication weakly increases coordination on Nash equilibrium outcomes, although average payoffs sometimes decrease. Twoway communication further improves payoffs in some games, but is detrimental in others. Moving beyond the class of symmetric 2 2 games, we nd that communication facilitates coordination in common interest games with positive spillovers and strategic complementarities, but there are also games in which any type of communication hampers coordination. JEL: C72. Keywords: Pre-play communication, cheap talk, coordination, level-k, cognitive hierarchy.Some people nd themselves in a new strategic situation. How can they best coordinate their actions? Since they cannot rely on precedence, maybe they should start talking? If so, what are the exact reasons why communication helps? These fundamental questions crop up in many disciplines, including evolutionary biology, psychology, political science, and economics. 1 Farrell (1987, 1988) and Matthew Rabin (1990Rabin ( , 1994 provide formal analyses of costless communication, or cheap talk, as a means to convey intentions and thereby improve coordination among rational players in games with complete information. 2 While the models are insightful, we argue that they make problematic assumptions concerning players' beliefs, and that recent models of strategic thinking offer alternative assumptions that better t our intuitions and the available experimental evidence.To put our arguments into perspective, let us brie y review some of the literature. Farrell (1987) studies communication in a Battle of the Sexes game (Figure 1). Farrell assumes that behavior will correspond to the symmetric mixed strategy Nash equilibrium if players cannot communicate. He also assumes that message pairs ."H ";"L"/ and ."L";"H "/ that are consistent with a pure strategy equilibrium will induce play of that equilibrium. Based on these assumptions, he shows that with two-way communication there are better symmetric mixed strategy equilibria than the no-communication equilibrium. Payoffs improve with the number of communication Ellingsen: