The idea that communication is a form of conveying contents through mind-reading goes back at least to Locke and, with important modifications, continues to be the dominant model through Grice and Kaplan. However, whereas Locke thought communication is entirely a matter of encoding and decoding content through linguistic forms, currently it is almost universally accepted, as Grice and Kaplan would have it, that meaning is at least partly determined through non-linguistic features of utterances. The idea that communication is trafficking in informational content has recently come under attack from (neo-)expressivists who argue that the role of a large fragment of natural language is not to communicate content. There have been two severe consequences of this shift: one is that modal vocabulary does not express informational content and the other is that classical patterns of inference like modus ponens and modus tollens are invalid. I will argue that these radical conclusions have a common source in the presumption of the Gricean and Kaplanean model of communication that meaning is partly determined by non-linguistic features of utterances. I will offer an account, Lockean in spirit, which explains the problematic behavior of modal expressions that motivated the radical conclusions, yet preserves the validity of classical patterns of inference, and the idea that modal vocabulary expresses informational content.