Firms routinely engage in public communications that are available to various constituencies, including competitors. In a laboratory experiment with prisoner's dilemma payoffs, the authors investigate the effect of one form of these communications - cheap talk signals: statements that are costless, nonbinding, and nonveriflable and do not directly affect the payoffs for either party. The authors find that only competitors that perceive that they share goals for a joint, coordinated outcome correctly update their beliefs about their competitor's next move on the basis of cheap talk signals. The authors contend that the conditions for cheap talk to work may be so rare that cheap talk is more likely to fall on deaf ears than to result in collusion. The authors suggest implications for managers and public policymakers as well as areas for further research.