In many intrastate armed conflicts civilians are the direct targets of violence by both governments and rebel groups; nonetheless, no quantitative study has ever examined and compared the determinants for government and rebel violence. I explain government and rebel attacks on civilians as violent bargaining strategies aimed at improving the bargaining position, and these strategies are dependent on the intensity level of the conflict. I propose that when fighting is low governments try to avoid killing civilians unless the threat is large enough, and rebels kill civilians to signal resolve in order to gain concessions. However, as the intensity level increases control becomes more important, so both parties target civilians to establish territorial control and undermine the support of the opponent. Using new data on killings of civilians I examine all conflict actors in an internal armed conflict, 1992 to 2004. The findings suggest that rebels use violence for communicative purposes in less intense conflicts, characterized e.g. by more violence when rebels are relatively strong and early in the conflict. In more intense conflicts, on the other hand, violence is used to secure control and compensate for lack in military capacity – then the weaker groups kill more civilians, and they are likely to kill more civilians the longer the conflict. Governments kill more civilians when the rebel opposition is strong; surprisingly they kill fewer civilians the longer the conflict, and democracy is not found to have any effect on government behavior.