This study investigated the dynamics of communication between members of an emergency management team faced with controlling a hazardous chemical spill. The effects of two sets of factors on communication were investigated; task-specific factors pertaining to characteristics of the emergency management task that are constant across different emergency situations, and situationspecific factors pertaining to the unique characteristics of the current situation. The results showed that both these factors were important in determining the pattern of communication between key team members. Verbal exchanges were found both to be correlated with the occurrence of critical events and to follow a 30-minute temporal cycle. The implications of the results for theories of naturalistic decision making are discussed.In emergency management, individuals must work together to contain and resolve a hazardous event. With each team member performing a specialized function, the effectiveness of the team depends upon the ability of its members to communicate with each other to coordinate activities, to share information, and to implement appropriate strategies (Caldwell, 1997). Decision making is thus said to be distributed across the members of the team (Brehmer, 1991; Rogalski and Samuray, 1993). But without a high level of interaction, distributed decision making will fail. For this reason, a critical factor affecting team function concerns the nature and timing of verbal exchanges between team members.The present study represents a relatively new approach to naturalistic decision making. Previous research has often been concerned with analyses of the cognitive aspects of decision making, based on the view that the most important determinants of behaviour are the cognitive representations of team members. Our study was based on a different view, namely that many important determinants of behaviour reflect characteristics of the task environment, the problems it poses, and the constraints it provides. On this view, theoretically important relationships exist between features of the task environment and human performance. To illustrate, Vortac et al. (1993) examined the behaviour of air traffic controllers in a simulation involving different degrees of traffic complexity. The onsets of various activities (e.g. issuing commands to aircraft) were recorded on-line, and the transition probabilities between different activities were analysed statistically. This analysis revealed a number of dynamical regularities that Vortac et al. (1994) explained