The growing prominence of the patent system has revealed an institutional structure in need of repair. Legal scholars and commentators have identified and discussed some of the difficulties of the individual institutions within the patent system. But they have not incorporated into their analyses a recognition of all of the relevant institutions and their highly interdependent roles. Because of this interdependence, patent reform requires multi-institutional analysis. This Article argues that the ostensibly discrete problems of our patent institutions stem from an initial mistake in institutional design. This mistake was made in 1982, when Congress implemented patent reform by attempting to concentrate expertise at the appellate level. Congress focused on the appellate courts even though the patent questions that require narrowly focused expertise are largely questions of fact, not law or policy. Multi-institutional reform must therefore focus on remedying the system's substantial deficiencies in the areas of both fact-finding and policymaking. Under a traditional administrative law model, primary responsibility for questions of fact would rest with the PTO. The role of the courts - whether the trial court or the appellate court - would be secondary. With respect to fact-finding, adopting certain aspects of the traditional administrative model is clearly necessary. Thus Congress should advance the PTO beyond its nineteenth century moorings and endow the agency with the expertise and neutrality necessary for it properly to wield significant fact-finding responsibility. At the same time, however, giving plenary responsibility on factual questions to the PTO would not be cost-effective. Accordingly, the trial courts should also be endowed with fact-finding expertise. Solving the allocation puzzle for questions of policy is more challenging. Each of the available institutional options - the legislature, the PTO, and the courts - has substantial associated liabilities. From the standpoint of political economy and institutional flexibility, the most disadvantaged institutional actor is probably Congress. The institutional choice thus comes down to one between the PTO and the appellate courts. On first examination, given the Federal Circuit's apparent aversion to policy analysis, its possible tendency towards pro-patent bias, and the well-known difficulties courts face in undertaking policymaking, the PTO might appear the better option. In order for the PTO to be a reasonably good policymaker, however, Congress would have to change the PTO into an altogether different agency: merely implementing reform that improved the agency's fact-finding capabilities would not be sufficient. In contrast, the level of change required in the court system would be less substantial. What we would need is greater generalist input in the appellate process. To some extent, the Supreme Court's increased interest in reviewing patent cases has already taken us down that path.