How does conflict exposure affect trust? We hypothesize that direct (firsthand) experience with conflict induces parochialism: trust towards out-groups worsens, but trust towards in-groups, owing to positive experiences of kin solidarity, may improve. Indirect exposure to conflict through third-party accounts, on the other hand, reduces trust toward everyone, owing to negativity bias. We find consistent support for our hypotheses in a lab-in-the-field experiment in Maluku, Indonesia, which witnessed a salient Christian-Muslim conflict during 1999-2002, as well as in three cross-country datasets exploiting temporal and spatial variation in exposure to violence. Our results help resolve a seeming contradiction in the literature and inform policies on resolving conflicts.