My dissertation investigates the factors that explain when groups are most or least likely to target civilians. While civilians are a particularly appealing target because they are easy to attack, many groups go to great lengths to avoid them. Thus, the question motivating my research is: given the potential windfalls implicit in civilian attacks, why do some organizations target them while others do not?. In this project, I take a two-step approach to answering this question. I begin by looking at what targets groups are likely to attack given the government from which they wish to gain political concessions. Based on a coercive model of violence, I argue groups attack those segments of society that apply the most pressure on a government. In general terms, the most influential segments of society for any government are its minimum winning coalition and selectorate. After surveying the literature on a large number of different regime types, I conclude that civilians face the highest risk of being attacked in democracies because, unlike forms of non- democracy, the minimum winning coalition and selectorate in democracies are composed of civilians. The second step, and the bulk of this project, examines civilian targeting within democratic contexts. To address variation within democracies, I argue that violent groups are most likely to attack civilians when doing so does not hurt their supporters or undermine the support they receive from the larger population. Although attacks on civilians are easy to execute and democratic political institutions encourage such attacks, these attacks impose high costs (often in the form of government repression) on the segment of the civilian population that supports the violent group. In addition, attacks on civilian supporters may alienate some of the group's supporters who will not back a group that targets innocent civilians. Losing civilian supporters is problematic for violent organizations that draw heavily from community-level alliances for operational success and political leverage. My theory identifies and explores three unique conditions that make groups more or less likely to attack civilians within democracies. First, when a violent group runs for political office, a community can punish the group. Consequently, groups with political wings temper their violence by attacking civilians less often. Secondly, when a terrorist group provides goods and services throughout neighborhoods for its supporters, it offsets the costs associated with violence on civilian targets. Compared to groups that don't distribute resources, provider groups are more likely to target civilians because the consequences are not as dire. Finally, when a terrorist group draws support from a highly segmented part of the population, it is easier for the group's enemies (usually government authorities, but a group's enemies may also include other violent groups) to repress and intimidate the group's supporters. When this occurs, groups are less likely to attack civilians because their support base suffers more costs from reprisals. Each condition emphasizes the same innovative contribution: civilians matter, and not just as collateral damage. They matter because violent groups rely on them and, as a result, they affect the type of violence groups pursue. I test my theory using large-N statistical analyses and detailed case research. My empirical work presents new data on terrorist groups, including their political (non- violent) and redistributive activities. Using this data, I test my claims using a cross-national dataset of attacks over the last forty years. My case research focuses on the Provisional Irish Republican Army (IRA) in Northern Ireland. The case of the IRA demonstrates how involvement in politics decreases the odds a group will attack civilians