Objective: Teamwork has been suggested as a promising approach to improving care processes in emergency departments (ED). However, for teamwork to yield expected results, implementation must involve behavior changes. The aim of this study is to use behavior analysis to qualitatively examine how teamwork plays out in practice and to understand eventual discrepancies between planned and actual behaviors.Methods: The study was set in a Swedish university hospital ED during the initial phase of implementation of teamwork. The intervention focused on changing the environment and redesigning the work process to enable teamwork. Each team was responsible for entire care episodes, i.e. from patient arrival to discharge from the ED. Data was collected through 3 days of observations structured around an observation scheme. Behavior analysis was used to pinpoint key teamwork behaviors for consistent implementation of teamwork and to analyze the contingencies that decreased or increased the likelihood of these behaviors.Results: We found a great discrepancy between the planned and the observed teamwork processes. 60% of the 44 team patients observed were handled solely by the appointed team members. Only 36% of the observed patient care processes started according to the description in the planned teamwork process, that is, with taking patient history together. Beside this behavior, meeting in a defined team room and communicating with team members were shown to be essential for the consistent implementation of teamwork. Factors that decreased the likelihood of these key behaviors included waiting for other team members or having trouble locating each other. Getting work done without delay and having an overview of the patient care process increased team behaviors. Moreover, explicit instructions on when team members should interact and communicate increased adherence to the planned process.Conclusions: This study illustrates how behavior analysis can be used to understand discrepancies between planned and observed behaviors. By examining the contextual conditions that may influence behaviors, improvements in implementation strategies can be suggested. Thereby, the adherence to a planned intervention can be improved, and/or revisions of the intervention be suggested.