Standard understandings of the separation of powers begin with the concept of function. Professor Nourse argues that function alone cannot predict important changes in structural incentives and thus serves as a poor proxy for assessing real risks to governmental structure. To illustrate this point, the Article returns to proposals considered at the Constitutional Convention and considers difficult contemporary cases such as Morrison v. Olson, Clinton v. Jones, and the Supreme Court’s more recent federalism decisions. In each instance, function appears to steer us wrong because it fails to understand separation of powers questions as ones of structural incentive and political relationship. In order to move away from function as the sole proxy for structural risk, the Article suggests a “vertical” approach toward separation of powers questions. That approach reconceives departmental power less as the power to perform a set of tasks fitting a particular constitutional description (e.g., adjudication, execution, legislation) than as a set of constitutionally created political relationships between the people and those who govern them. Put another way, the separation of powers becomes less a search for transcendental descriptions of the departments than a means of considering how shifting structure affects liberty—how structural incentives may incline governmental actors to act toward the people in ways that risk the electoral powers of both majorities and minorities. † Victoria Nourse, Associate Professor of Law, University of Wisconsin. Special thanks to Ann Althouse, Neil Komesar, and Jane Schacter for reading early drafts of this Article, as well as to the members of the Columbia Legal Theory Workshop and, in particular, to Michael Dorf, George Fletcher, and Peter Strauss for listening to the some of the ideas presented in an early version of this Article. I would also like to thank the students in Professor Komesar’s Law and Economics seminar, who read and commented on a later draft. The editors of the Duke Law Journal, in particular Scott Thompson, provided excellent assistance. All errors are, of course, my own. NOURSE TO PRINTER 06/09/00 3:26 PM 750 DUKE LAW JOURNAL [Vol. 49:749