Religious conflict has increased in scope and intensity in recent decades. By drawing on new data, theories, and methodologies, this dissertation advances the political science literature on religion and armed conflict in three key ways.

First, the dissertation uses a new dataset on Religious Claims and Recruitment (RCR) to explain variation in civilian targeting during civil wars. It proposes both ''Defender of the Faith'' and ''Religious Revolutionary'' pathways linking religion to one-sided violence, and finds strong evidence that religious militants target civilians more extensively than non-religious ones. Further, it also finds strong empirical support for the ``Religious Revolutionary'' pathway in particular.

Second, the dissertation is the first to apply nonparametric machine learning to the study of foreign fighter recruitment. Why were some countries more likely than others to produce Syrian foreign fighter contingents? By using Bayesian Additive Regression Trees (BART), the dissertation shows that the extent to which a country debated, introduced, or enforced a nationwide restriction on Islamic headdress in the years prior to the onset of the Syrian civil war is an important predictor of subsequent foreign fighter outflows. Case studies from Belgium and Quebec extend the finding further.

Third, the dissertation develops a novel theory for why some religious organizations resort to violence, while others engage in peaceful protest. The theory holds that an organization's political theology uniquely constrains its political behavior, with organizations that hold exclusionary theologies being most likely to resort to violence. A key advantage of this theory is that it generalizes across religions. Paired case studies from the post-war United States and contemporary Mali and Tunisia confirm the theory's observable implications.