The effects of task importance and group decision training on the discussion behavior of decisionmaking groups were investigated. Three-person groups decided which of 3 hypothetical faculty candidates would be the best person to teach an introductory psychology course. Prior to discussion, some of the information about each candidate was given to all group members (shared information), whereas the remainder was randomly divided among them (unshared information). In general, groups discussed much more of their shared information than their unshared information. Increasing the importance of the task slowed the rate at which information was brought forth during discussion. By contrast, group decision training increased the amount of both shared and unshared information discussed and altered the sequential flow of shared and unshared information into the discussion: Discussion in untrained groups focused first on shared information and then on unshared information; discussion in trained groups did not shift focus over time. Results are discussed in terms of an information-sampling model of group discussion and the role of discussion in group decision-making effectiveness.Decisions about important social, organizational, and political issues are frequently made by groups rather than individuals. Using groups to make decisions is often justified on the grounds that groups can bring more intellectual resources to bear on a problem, which in turn should increase the probability that a high-quality decision will result (e.g., Vroom & Jago, 1988). One such resource is the diverse store of knowledge held by group members. Because of differences in background and experience, group members frequently have different information about the choice alternatives under consideration. To exploit this knowledge, decision-making groups usually attempt to pool the unique information their members hold. This is typically done through face-to-face discussion, though other alternatives exist (e.g.,