Can police be trained to treat people in fair and respectful ways, and if so, will this influence evaluations of the police and crime? To answer these questions, we randomly allocated 120 crime hot spots to a procedural justice (PJ) and standard condition (SC) in three cities. Twenty-eight officers were randomly assigned to the conditions. The PJ condition officers received an intensive 5-d training course in the components of PJ (giving voice, showing neutrality, treating people with respect, and evidencing trustworthy motives). We used police self-report surveys to assess whether the training influenced attitudes, systematic social observations to examine impacts on police behavior in the field, and arrests to assess law enforcement actions. We conducted pre and post household surveys to assess resident attitudes toward the police. Impacts on crime were measured using crime incident and citizen-initiated crime call data. The training led to increased knowledge about PJ and more procedurally just behavior in the field as compared with the SC condition. At the same time, PJ officers made many fewer arrests than SC officers. Residents of the PJ hot spots were significantly less likely to perceive police as harassing or using unnecessary force, though we did not find significant differences between the PJ and SC hot spots in perceptions of PJ and police legitimacy. We found a significant relative 14% decline in crime incidents in the PJ hot spots during the experiment.