In his later writings Quine is increasingly explicit about the fact that his view of language is, in a certain sense, structuralistic. Structuralist interpretations of non-empirical, especially mathematical theories are now commonplace, but this chapter argues that Quine’s thought experiment with radical translation can be interpreted as showing that even empirical theories cannot be anchored in reality so firmly as to evade the same structuralist nature. Therefore, this peculiar form of structuralism extends to all our theories––the terms of all of them are best seen as meaning not definite substances, but nodes in certain structures. Moreover, radical translation shows––or purports to show––that the structure behind any natural language allows for some non-trivial ‘automorphisms’––that mapping the meaning of rabbit on that of undetached rabbit part, provided we make an appropriate remapping of many other meanings, does not change the language. Inscrutability of reference is then only a direct consequence.