Why does repression sometimes work to stop violent protest and sometimes heighten protest? We argue that the effect of repression on protest depends critically on a "memory of violence" within the state. Without this memory, the costs of continued protest in the face of increased repression are often too great for unrelenting mobilization, effectively suppressing the political violence. We focus on a global sample of repression and protest data at the weekly level from 1990 to 2009. In states with civil war histories, repression can mobilize a population previously primed for violent protest.