We often treat earlier decisions as constraining later ones, even when the arguments justifying them differ in structure. When, and in what way, do more complex arguments constrain, or defeat, simpler ones? Building on a hierarchical model of precedent-based reasoning from AI and Law, I compare two strategies for modeling these constraints—the flattening strategy and the derived-priority strategy—and show that the former fails in two systematic respects. Based on this analysis, I formulate two rules that implement the latter.