Organizations increasingly rely on teams to respond to crises. While research on team effectiveness during nonroutine events is growing, naturalistic studies examining team behaviors during crises are relatively scarce. Furthermore, the relevant literature offers competing theoretical rationales concerning effective team response to crises. In this article, the authors investigate whether high- versus average-performing teams can be distinguished on the basis of the number and complexity of their interaction patterns. Using behavioral observation methodology, the authors coded the discrete verbal and nonverbal behaviors of 14 nuclear power plant control room crews as they responded to a simulated crisis. Pattern detection software revealed systematic differences among crews in their patterns of interaction. Mean comparisons and discriminant function analysis indicated that higher performing crews exhibited fewer, shorter, and less complex interaction patterns. These results illustrate the limitations of standardized response patterns and highlight the importance of team adaptability. Implications for future research and for team training are included.