Debates over whether a certain thing is (or ought to be) construed as 'political' are frequent and frequently interminable. This article argues that approaches to the proper understanding, scope and application of political concepts should recognise that they are both normative and contestable and also that, because they are employed by both theorist and theorised, they can never be sharply defined. It is argued that many debates achieve no theoretical closure because the terms of discussion are confined by a certain understanding of concepts as empirical and classificatory. This article examines these issues by using the work of R.G. Collingwood to suggest that conceptual overlap is inevitable and also that the theoretical analysis of politics should distinguish between the empirical and the theoretical phase of the concept. Philosophically, politics is not a separate sphere of activity but a dimension of all activity, and the correct way to understand politics is to understand it as activity, not as substance. For certain empirical purposes we categorise some things as political and others as non-political, but in doing so we should be careful whether we are doing so philosophically, historically or through stipulative definition. This article does not seek to cover all ramifications of the debate or its later literature, but to suggest that Collingwood's approach has something to contribute to the analysis of political concepts. © Political Studies Association, 2005.