In this paper we challenge the key assumption in discussions of racial "profiling": that the problem is mainly a matter of how individual officials think and who they choose to stop. Certainly the thinking processes of individuals are one element of the process that produces racial disparities. But an equally important element is how the practices of front-line workers are organized institutionally. Using narrative and survey data from a large survey of drivers, we show that racial disparities in who is stopped by the police and what happens during the stop appear in patterns that are best understood as reflections of institutionalized practices. These practices are organized through professional training, disciplinary expectations, and shared professional lore, and are to a considerable degree deliberately crafted by institutional leaders; they are not merely individual choices or the product of inchoate cultural expectations. These institutional practices do set the stage for the expression of widely shared negative stereotypes of black criminality. We provide evidence for this thesis in examples drawn from police training manuals. We also show that patterns in who is stopped by the police grow from these institutionalized practices. For example, not every officer all of the time selects black drivers for stops at higher rates than whites; instead, these disparities appear in predictable patterns that trace the organization of police stop practices. We also show that while myriad personal and vehicle characteristics may affect the risk of being stopped, one stands out as overwhelmingly more important than the rest: the person's race.