A major characteristic of the linguistic theories of the twentieth century which stemmed from logical empiricism is that the notion of reference was placed at the core of semantics. As a result, the meaning of a noun was associated with the object or the entity to which it referred, and the meaning of a statement was correlated with the state of affairs or the possible world it was meant to designate. This representational conception, which corresponds to the definition of semantics proposed by Charles Morris in 1938 (semantics studies “the relations of signs to the objects to which the signs are applicable”), became, however, the target of a number of criticisms at the end of the last century. Linguists objected that it only reduces language to a world that does not correspond to the semantic interpretation performed by the speakers. In his book Making It Explicit (1994), Robert Brandom argued that this representational conception had to be replaced by one which restores the value of the notion of inference. But, for the semantic explanation, by proposing such a reversal of perspective in prioritizing the concepts of “reference” and “inference”, the respective role and scope of investigation of semantics and pragmatics were deeply modified. Therefore, the aim of this communication will be to retrace the evolution of the connections between semantics and pragmatics by comparing some representational and inferential approaches. First, we will analyse the referential conception of meaning and we will show how a semantics based on a theory of models can be open to what Wilfrid Sellars (1956) and Willard Quine (1960) have respectively called the “myths of the Given and of the Museum”. Then we will examine the inferential approach to meaning by underlining the various theoretical changes in orientation it induces (primacy of the proposition over subsentential expressions, holistic conception, ...). Finally, this analysis will lead to highlighting some consequences of the modification of the role of semantics and pragmatics. More specifically, we will stress the fact that certain parameters, which in a referential conception belonged to pragmatics, now belong to semantics, while from an inferential perspective, the taking into account of the referential dimension of languages is understood as a social practice which is part of pragmatics.