This paper reviews contending realist assumptions about domestic and systemic impulses for balancing behavior, derives a set of corresponding hypotheses for state actions and submits them to a statistical large-n analysis for testing. A total of 18 highly conflict-prone dyads of states are observed over lengthy periods of time in order to gather data for a regression analysis of the effects of different impulses on both the external and internal balancing behavior of the weaker states. In accordance with the results, it is argued that domestic (or unit-level) factors are highly important in explaining the scope of balancing and often exert a stronger influence than do power gaps between states. As moderating factors, they are especially crucial in clarifying apparent cases of over-and underbalancing.