"Discourse history" can be conceived of as a sub- or 'sister'-discipline of historical linguistics. But how far does such theory construction help us to establish a coherent set of objects and methods of empirical investigation? In which way can modern corpus-based/driven research be integrated with qualitative analyses of semantic and stylistic change? The paper will discuss these questions with regard to competing accounts of metaphor development, specifically concerning the conceptualization of STATE/SOCIETY as a (HUMAN) BODY. Aspects of this metaphorical mapping have become lexicalized as set phrases (body politic, head of government etc.) that are still in use, and we can trace them back to medieval times and even to antiquity. However, does such a tradition of use constitute a "discourse history" in an empirically testable sense? Or are we only dealing with repeated instances of a fundamentally 'ahistorical' cognitive operation? The paper concludes with a differentiation of levels of diachronic analysis that relates the cognitive analysis of "conceptual metaphors" to the modeling of their development in historical discourse traditions. (Andreas Musolff)