Rebel attacks on civilians constitute one of the gravest threats to human security in contemporary armed conflicts. But why do rebel groups kill civilians? The dissertation approaches this question from a strategic perspective, trying to understand when and why rebel groups are likely to target civilians as a conflict strategy. It combines quantitative studies using global data on rebel group violence with a case study of the civil war in Mozambique. The overall argument is that rebel groups target civilians as a way of improving their bargaining position in the war relative to the government. The dissertation consists of an introduction, which situates the study in a wider context, and four papers that all deal with different aspects of the overall research question. Paper I introduces new data on one-sided violence against civilians, presenting trends over time and comparing types of actors and conflicts. Paper II argues that democratic governments are particularly vulnerable to rebel attacks on civilians, since they are dependent on the population. Corroborating this claim, statistical evidence shows that rebels indeed kill more civilians when fighting a democratic government. Paper III argues that rebels target civilians more when losing on the battlefield, as a method of raising the costs for the government to continue fighting. A statistical analysis employing monthly data on battle outcomes and rebel violence, supports this argument. Paper IV takes a closer look at the case of Mozambique, arguing that the rebel group Renamo used large-scale violence in areas dominated by government constituents as a means for hurting the government. Taken together, these findings suggest that violence against civilians should be understood as a strategy, rather than a consequence, of war.