Source: (1996) Paper presented at the conference 'Problem-Solving Policing as Crime Prevention'. Stockholm, 1996.Policing domestic violence is the most scientifically developed application of the problem-solving paradigm. Its current state reveals many basic challenges to the paradigm as a tool for crime prevention. While there has been great progress in adopting the goal of preventing domestic violence, there has been less understanding and use of the paradigm's methods for accomplishing that goal. Almost no work has been devoted to defining the problems more precisely (classification). Analyses of risk factors have been approached with basic errors of logic (prediction). Evaluations of police responses, while scientifically rigorous, have been unwelcome due in part to their complexity (causation). In the short run, both violence prevention and science can be fostered by better measurement of domestic violence trends. In the long run, progress in identifying more effective policing strategies for prevention and control can be achieved only by a culture of science in policing, just as medicine was transformed by such a cultural change in the early Twentieth Century, in which trial and error becomes the norm for crime prevention.