Conventional wisdom holds that war inevitably erodes trust. We challenge this notion and argue that the link between violent conflict and trust is more complex. There are in fact two faces of war and trust: While trust in strangers and people different from ourselves is reduced by civil war, this is not the case for the trust we place in most people in our immediate social surroundings, where trust may even increase in the wake of war and destruction. We lay out a theoretical framework that unravels the underlying mechanisms of how civil wars influence different dimensions of social trust and subject these predictions to an empirical test, combining survey data with battle deaths data in a unique way. Our empirical findings based on hierarchical linear models suggest that people living in countries that experienced civil wars with a high number of casualties indeed show less generalized trust. However, in accordance with our theoretical model, the intensity of civil wars is not systematically related to particularized trust.