Earlier philosophers who analyzed the role of fundamental theories in physics generally assigned linguistic analysis a preliminary role. Quine and Sellars each developed methodologies for proceeding from an ordinary language framework to a final unifying framework based on anticipations of the structure ultimate theories should have. The assumption was that only the final framework supplies the true ontology. Reactions against their syntactic conception of theories led to a semantic conception of theories. The emphasis on individual theories led to interpretation through models with no role for linguistic analysis. However, linguistic considerations are indispensable when one analyzes the normal functioning of experimental and theoretical physics. Ordinary language implicitly contains various inference-supporting structures that were gradually modified in the protracted development of the language of physics. Here we consider two such structures, categorization and the metaphorical extension of terms to new contexts. We also draw on Davidson's semantics.