The actual-language relation is the relation between a speaker S and a language L and in virtue of which L is *the* language of L. I present Lewis' (1975) conventional account of the actual-language relation and the meaning-without-use objection raised against it by Schiffer and Hawthorne. The MWU-problem shows that the actual-language question cannot be separated from the question about how to justify assumptions about subsentential structure (grammar). I argue that Lewis' solution to the problem is flawed and sketch a more promising approach.