There are two approaches to community policing, one that expects to reduce the frequency and intensity of coercion, and another that encourages coercion at a lower intensity. Police coercion in two departments, one that embraces each approach, were analyzed. Indianapolis, IN, and St. Petersburg, FL, were the research sites for the Project on Policing Neighborhoods (POPN), the study from which the data were drawn. The study used two data sets from the POPN: systematic observation of patrol officers and in-person interviews of those officers. Coercion was defined as acts that threaten or inflict physical harm on citizens. Community policing is measured by the department leadership’s dominant model, patrol officer job responsibility, amount of training in community policing principles, and the officer’s personal attachment to each of the two models of community policing. Results show that the model of community policing that top management promotes does seem to matter. Officers at the broken windows site (Indianapolis) showed a striking proclivity for greater coercion than did those at St. Petersburg. There was little support for officer-level influences and no support for the notion that either model increased verbal coercion while decreasing physical coercion. Indianapolis officers were less positively disposed to aggressive patrol than St. Petersburg officers were, although the difference was not statistically significant. They also averaged a substantially more positive attitude toward community partnership than did the St. Petersburg officers. The effects of the leader’s approach to community policing depend on the model employed. There are great differences among approaches that pass as community policing, and these differences have important consequences for what happens on the street. The officers’ attitudes toward the two community policing models showed no effect at either site. 6 tables, 55 citations