Why do insurgencies erupt in some places and not in others? This article exploits an original violent event database of 274,428 municipality-month observations in Colombia to determine the conditions favoring organized violence at the subnational level. The data cast doubt on the conventional correlates of war: poverty, rough terrain, lootable natural resources, and large, sparsely distributed populations. The evidence suggests that rebellions begin not in localities that afford sanctuaries, impoverished recruits, and abundant finances, but instead in regions providing receptacles of collective action: the organizational legacies of war. Specifically, the data indicate that regions affected by past mobilization are six times more likely to experience rebellion than those without a tradition of armed organized action. The significant correlation between prior and future mobilization is robust across different measurements of the concepts, levels of aggregations of the data, units of analysis, and specifications of the model. These include rare events and spatial lag analyses. These results highlight the need for micro conflict data, theory disentangling the causes of war onset from those of war recurrence, and a reorientation away from physical geography and back to the human and social geography that determines if rebellion is organizationally feasible. The findings suggest new avenues of research on the post-war trajectories of armed organizations, the causes of repeated war, and the micro-foundations of rebellion.