Much of the recent literature on freedom of speech has focused on the arguments for and against the regulation of certain kinds of speech, such as hate speech and pornography. Less attention, however, has been paid, at least recently, to the analysis of the very normative foundations of freedom of speech. This is overwhelming true of the legal literature on free speech, but it is also a noteworthy aspect of the recent philosophical work which we might expect to be more concerned with fundamental questions. This paper makes a fresh start on the difficult task of setting out the normative foundations of freedom of speech. More specifically, we adopt a new approach, one that ties the limits to free speech to those very values that free speech serves. The conceptual innovation which enables us to set out a unified account of the nature and limits of free speech is to theorise free speech as a relational, rather than an individual, value. By drawing on the conceptual and normative resources of recognition theory and republican political theory, we argue that free speech consists of a relation between the speaker and receiver of speech; where that relation is valuable, speech should be protected; where it is not, it may be restricted. In the first part, we outline in more detail the problems we see in current philosophical approaches to the issue free speech. Those problems arise from their failure to meet five basic desiderata which, we maintain, an adequate theory of free speech has to meet. In the second part, we set out our key argument in some depth and develop our theory. In the third and final part we apply the theory to the analysis of hate speech.