There is demand for police reform in the United States to reduce use of force and bias, and to improve police-citizen relationships. Many believe de-escalation should be a more central feature of police training and practice. It is suggested that improving officers’ communication and conflict resolution skills will temper police-citizen interactions and reduce police use of force, and that such a change will improve citizen trust in the police. To date, however, de-escalation training has not spread widely across agencies, and de-escalation as a strategy has not been studied. Without an evidence-based understanding of these concepts, de-escalation training will proceed blindly, if at all. Accordingly, this dissertation represents one of the first empirical studies of de-escalation in police work. The author completed this study as an embedded researcher in the Spokane (WA) Police Department, and it proceeds in two parts. Part 1 was exploratory and qualitative, consisting of in-depth interviews (N=8) and a focus group (N=1) with eight highly skilled police de-escalators. These officers were nominated by peers as the best among them at de-escalating difficult encounters with citizens. The results in Part 1 explore officers’ perceptions of de-escalation and offer a definition of de-escalation as well as a description of de-escalation tactics. In Part 2, the author systematically observed the concepts developed in part 1 during 35 ride-alongs with 29 police officers, including the peer nominated officers (N=131 police-citizen encounters). This phase of the research investigated whether characteristics of officers, citizens, and situations are associated with de-escalation use, and de-escalation effectiveness. Implications from these findings are drawn for police practice, theory, and research methods. This dissertation is a launching point for empirical research on de-escalation in police work.