This paper addresses the endogeneity of ethnic settlement patterns and conflict, that is, how settlement patterns affect conflict, and how conflict in turn changes the ethnic map. I argue that the application of violence during conflict is driven by the territorial aspirations of ethnic groups. Locations where territorial claims clash should see more violence as groups struggle for control of a unit. More precisely, in an attempt to secure control over these locations, there should be more violent confrontations between the group’s military forces. For the same reason, these locations should also experience more one-sided violence against civilians. The effect of conflict on territory should be such that by means of moving populations, it decreases the level of contestation across all units. I study the dynamics of group geography and conflict in Bosnia using data on ethnic population shares at the municipality level, both from before and after the war. These data are combined with information on conflict events from the Armed Conflict Location and Events Dataset (ACLED). I construct a spatial indicator measuring the degree to which the territorial claims of ethnic groups clash at a particular location. I find support for higher levels of violence at these contested locations. Furthermore, post-war contestation scores are significantly lower, which points to a pattern of strategic ethnic unmixing during conflict. However, my results only partly support the impact of local violence as a trigger of this unmixing.