The functionalist viewpoint in linguistics can take different forms. A caricature of functionalist thinking is the notion that the structure of language is optimised, or nearly so, for its function as a means of human communication. This notion has met with widespread scepticism because of its lack of predictiveness in the face of typological variation. Either it leads to the prediction that all languages are en route to some single ‘Utopian’ (even if they have not quite achieved it) or it leads one to posit so many contradictory functional goods that the nature of possible languages is not effectively restricted. A second, and far more sophisticated, understanding of functionalism is the claim that there are regular relations between the way language is represented in the mind and the way that it is processed during speech production and perception. These relations arise because language is acquired from experiences of use, and because even in adults patterns of use affect cognitive representations. The effects of individual instances of language use are local, incremental and context-dependent. Language use and competence in a language are thus two aspects of a single system. Multiple system configurations are possible for the same reason that multiple ecosystems are possible; like the products of biological evolution, human languages are merely good enough, and not globally or absolutely optimised. This understanding of linguistic functionalism has proved fruitful for at least two decades and is now coming into its own. Its rise constitutes part of the rise of scientific research on complex systems and emergent structures generally, in areas ranging from geophysics and granular media to population biology.