Many constitutions specify procedures that allow critical decisions to be made with a different rule from day-to-day decisions. We propose a theory of constitutional rules that explains why. The theory is based on the assumption that the type of decision can be observed, but not verified. We characterise two classes of second-best constitution, both with clear analogues in real world constitutions: i) incentive scheme (IS) constitutions that elicit information about the type of decision through costly decision rule switching procedures, and ii) linking scheme (LS) constitutions that grant limited veto powers to interested parties. We explore how the relative performance of the IS and LS constitution depends on the economic environment.