In this paper we discuss Brandom's attempt to analyse the working of language without taking the concept of representation as given. Three main topics are presented: the inferential definition of meaning, the criticism of epistemology and the anaphoric treatment of direct reference. The basic idea of meaning as inference is explained using the concepts of commitment and entitlement: to understand the meaning of a sentence is to understand which other sentences you are committed to recognize as true and on which ground you are entitled to assert the sentence. The criticism of naturalized epistemology is widely developed also in "Articulating Reason", a series of lectures which are a development of themes of the main book. We give here some hint on the contrast with Alvin Goldman on role of the causal theories of knowledge. The alternative to direct reference theory is discussed with an example on Kripke's distinction between speaker's reference and semantic reference. Even this distinction can be placed in a framework where anaphora is the leading concept which helps to explain the working of our language in communication. Eventually, some internal difficulties of Brandom's project are hinted at, even if they do not depriive the general framework of its interest.