Theories of conflict emphasize dyadic interaction, yet existing empirical studies of civil war focus largely on state attributes and pay little attention to nonstate antagonists. We recast civil war in a dyadic perspective, and consider how nonstate actor attributes and their relationship to the state influence conflict dynamics. We argue that strong rebels, who pose a military challenge to the government, are likely to lead to short wars and concessions. Conflicts where rebels seem weak can become prolonged if rebels can operate in the periphery so as to defy a government victory yet are not strong enough to extract concessions. Conflicts should be shorter when potential insurgents can rely on alternative political means to violence. We examine these hypotheses in a dyadic analysis of civil war duration and outcomes, using new data on nonstate actors and conflict attributes, finding support for many of our conjectures.