The distinction between grammaticality and acceptability has been regarded with strong skepticism in functional linguistics because of its origin in Chomskyan linguistics. In this paper I will argue that the distinction is useful in functional linguistics, provided that it is based on a distinction between competence and performance, rather than on a distinction between syntax and meaning. The basic rationale for having such a distinction is that much of linguistics is concerned with describing relatively stable grammatical knowledge, rather than the psycholinguistic dynamics of language use. The article will briefly summarize the early history and rationale of the notion of grammaticality within Chomskyan and functional linguistics, before defining a functional, usage-based definition of grammaticality. Finally, the article will illustrate how this usage-based notion of grammaticality can be used as a framework for interpreting corpus and experimental data on language use.