This article describes an agent-based simulation model of ethnonationalist radicalization between political actors and their constituencies based upon evidence from the former Yugoslavia. The central mechanism is the recursive feedback between political and cultural dynamics, focusing on processes prior to the outbreak of actual violence. The results offer theoretical insights by revealing mechanisms that lead to escalation. These can be found within politics as well as among the population: between conflicting ethnically homogeneous regions, opposing radicalization forces fuel an escalation spiral. These processes are driven by political influences. Challenging the theory that diversity breeds conflict, this suggests that multiethnic regions are more capable of withstanding political pressures. However, they are vulnerable to imported violence, driven by the local population. This finding is tested with a different model of the same events, in which different implemented mechanisms generate results, in line with the diversity-breeds-conflict theory. A comparative discussion demonstrates how simulation is sensitive to theoretical predispositions.