Abstract In reaction to high-profile incidents of excessive and deadly force, policymakers, advocates, scholars, and the general public, have all called for police departments to embrace de-escalation training as a method for improving police-citizen interactions. This practice has, in turn, spurred a small, but growing, number of evaluations of police de-escalation training programs. The findings of these studies have been mixed, but incomplete. In particular, we argue that prior studies of de-escalation have been hindered by (1) a lack of consideration of changes in officer behavior in incidents not involving force, (2) a singular focus on whether or not force was used rather than alterations to the “trajectory” of use-of-force encounters, and (3) a failure to measure the intervening mechanisms between de-escalation training and officer behaviors (i.