This paper investigates how variety affects the innovation output of a region. Borrowing arguments from theories of recombinant innovation, it is expected that related variety will enhance innovation as related technologies are more easily recombined into a new technology. However, it is also expected that unrelated variety enhances technological breakthroughs, since radical innovation often stems from connecting previously unrelated technologies opening up whole new functionalities and applications. Using patent data for US states in the period 1977–99 and associated citation data, evidence is found for both hypotheses. This study thus sheds a new and critical light on the related variety hypothesis in economic geography.