This dissertation examines two related questions: why do some minority groups rebel after having been granted territorial autonomy while others do not? If autonomy does not reduce ethnic rebellion, why do governments grant it in the first place? By developing a game-theoretic model, I argue that internal divisions within minority groups play an important role in determining which groups get autonomy and how much, and how the autonomy affects ethnic rebellion in the end. As for autonomy granting, within-group divisions do not reduce the chances that ethnic groups get autonomy, but internally divided groups receive smaller amounts of autonomy than unitary ones. As for ethnic rebellion, the effect of territorial autonomy is conditional on the internal structures of ethnic groups. I argue that territorial autonomy is more likely to reduce the occurrence of ethnic rebellion only for more unified groups. However, for internally divided groups, territorial autonomy reduces the intensity of ethnic rebellion rather than its occurrence. Thus, governments use territorial autonomy not only as a means to achieve peace, but also as a means to select into small conflicts which yield higher payoffs than costly peace. In order to evaluate the theory, I conduct a quantitative study of self-determination disputes from 1985 to 2003. The analysis reveals that the internal structure of minority groups has a systemic effect on both autonomy settlements and conflict. Autonomy granting reduces the occurrence of large-scale ethnic rebellions rather than all rebellions.