This essay argues that the logical significance of most natural language expressions is indefinitely elastic. This, it is argued, undermines the idea that the meaning of a word is an item for which it stands, and puts pressure on the methods of conceptual analysis and theoretical elucidation that require context-invariant stable application conditions. Furthermore, it is argued that the insistence that such semantic content is needed which—impervious to local pragmatic concerns—remains stable and available for reasoning, gets things back to front. For in order to determine the correct inflection of any given use of an expression, its inferential relations—in the context of an utterance—must already be discerned. The lack of contextually independent content, however, presents no mystery about language use. For the ability to understand what is said is explained not only by shared practices and common interests, but also by the capacity for interlocutors to ask questions and explain what they mean.