Understanding how preferences are formed is a key question in the social sciences. The ability of agents to interact with each other is a prerequisite for well-functioning societies. Nevertheless, the process whereby the preferences of agents in conflict are formed have often been black boxed, and the literature on the effects of armed conflict on individuals reveals a great variation in terms of outcomes. Sometimes, individuals are willing to cooperate and interact even with former enemies, while sometimes, we see outright refusal to cooperate or interact at all. In this dissertation, I look at the role of threat in driving some of these divergent results. Armed conflict is rife with physical threats to life, limb and property, and there has been much research pointing to the impact of threat on preferences, attitudes and behavior. Research in the field of evolutionary psychology has revealed that threat is not a singular category, but a nuanced phenomenon, where different types of threat may lead to different responses. I argue that by taking a more nuanced approach to threat, drawing on theories from the field of evolutionary psychology, some of the variance in outcomes can be explained. In particular, some commonly observed features of protracted conflicts, such as seemingly indivisible issues and parochialism may be moderated by threat. In the four essays, I address this from both a theoretical and empirical point of view. In Essay I, I illustrate in a formal bargaining setting how threats can lead actors to prefer risky all-or-nothing gambles to division schemes, preventing bargaining solutions to be found. In the second essay (Essay II), I show that willingness to make compromises with members of other groups are not contingent on group affiliation alone, but rather on the expected reciprocity of that group. Furthermore, characteristics of others beyond group can also affect pro-sociality. Based on a threat management perspective, me and my coauthors in Essay III show that non-threatening social categories, such as women or the elderly, are shown higher levels of altruism, also when there individuals are outgroup members. Exposure to violence can even increase altruism across group lines, but only to these non-threatening groups. Finally, in Essay IV, I show that those who experience post-traumatic growth as a result of traumatic events reverse the standard loss aversion people generally display, even showing gain-seeking preferences. Together, these results point to the importance of bringing in a more nuanced conceptualization of the role of threat in the study of peace and conflict. Even in the most brutal and destructive conflicts, humans are able to cooperate, also across group lines. However, for this cooperation to function, managing the threats in conflict is of central importance.