There are many differing ways to be a realist about language. This paper seeks to classify some of these and to examine the implications of each for the study of language. The principle of classification it adopts is that we may distinguish between realisms on the basis of what exactly it is that they take to be real. Examining in turn realisms that ascribe reality to the external world in general, to causal mechanisms, to innate capacities, to linguistic signs, to social structures, to language systems, and to linguistic groups, the paper summarises the case for a particular critical realist ontology of language. In the process, it engages briefly with the work of Saussure, Chomsky, Halliday, and more recent explicitly realist thinkers such as Bhaskar, Pateman, Archer, Sealey and Carter. One implication is that language itself is not a phenomenon that separates us from a causally structured world, but rather a part of that world, a part with an identifiable causal structure of its own that is similar to that of other normative phenomena.