Despite a widespread trend toward the adoption of increasingly participatory approaches to political decision making in developing countries, there is little or no evidence that these practices in fact return the benefits attributed to them. This article investigates one specific worry—that participatory decision-making processes may be vulnerable to manipulation by elites. The authors report on a field experiment, drawing on a unique nationwide experiment in democratic deliberation in São Tomé and Príncipe in which the discussion leaders were randomly assigned across meetings. The randomization procedure provides a rare opportunity to identify the impact of leaders on the outcomes of group deliberations. They find that leader effects were extremely large, in many cases accounting for over one-third of all variation in the outcomes of the national discussions. These results have important implications for the design of such deliberative practices. While the total effect of leadership cannot be assessed, it may still be possible to observe when leader influence occurs and to correct for leader effects in comparisons of outcomes across deliberations.