The functionality doctrine serves a unique role in trademark law: unlike virtually every other doctrine, functionality can trump consumer confusion (or so it seems, at least in mechanical-functionality cases). In this sense, functionality may be the only doctrine in trademark law that can truly be considered a defense. But despite its potential power, the functionality doctrine is quite inconsistently applied. This is true of mechanical functionality cases because courts differ over the extent to which the doctrine focuses on competitors’ right to copy unpatented features as opposed to their need to copy. And aesthetic functionality cases are even more scattered: some courts refuse to recognize the aesthetic-functionality doctrine at all, and courts that do recognize it are often reluctant to actually find the features at issue functional, even when exclusive use of those features seems very likely to put competitors at a significant, non-reputation-related disadvantage. The problem is not simply that courts do not understand or do not like the functionality doctrine, though there is reason to believe both of those conclusions are warranted. It is instead that courts have fundamentally different views about the purposes of functionality. These differing views reflect a longstanding lack of consensus about trademark law’s proper role in competition policy and equally longstanding, if unexamined, intuitions about the types of features that are competitively important. Put simply, trademark law lacks a sufficiently robust theory of legitimate competition against which particular actions can be judged “unfair.” This Article uses functionality as a means of highlighting courts’ lack of consensus about the relationship between trademark law and competition.