The creation in 1982 of the United States Court of Appeals for the Federal Circuit represents the first significant appellate consolidation of a particular area of law in American history. Evaluating the Federal Circuit experiment is highly important to understanding, and perhaps improving upon, the institutional design of the federal judiciary. The Federal Circuit was grounded in a congressional desire for greater uniformity in the application of patent law. In patent law, as in other areas of the law, uniformity is a virtue. But uniformity is not the only virtue and centralization has its costs. The issue of centralization versus decentralization manifests itself in numerous areas of law, politics, economics and business. This article draws upon that literature and argues that the time is ripe to rethink the Federal Circuit experiment and the fixation on uniformity that gave rise to the experiment. The criticisms currently levied against the Federal Circuit - that it maintains excessive insularity, is subject to path dependency in its case law, and produces inadequately nuanced jurisprudence - can be traced back ultimately to the court's chief structural limitation: The court lacks the benefit of sister-circuit jurisprudence that would engender a healthy competition of rationales and provide a mechanism for testing legal innovations. Accordingly, the article proposes that in addition to the Federal Circuit, at least one, and perhaps two or three, extant circuit courts should have jurisdiction to hear appeals relating to patent law. This proposal represents a shift in strategy from one dominated by the pursuit of uniformity, to one where competition and diversity are equally important. As the literature from many other areas suggests, a choice between centralized and decentralized institutions cannot and should not be made with a polar solution. The issue is one of optimization. In 1982 Congress decided that the optimal number of federal appellate courts deciding patent cases was fewer than thirteen; we suggest that the optimal number may also be greater than one
