If philosophical semantics and the majority of grammars give a certain privilege to the mechanisms of the production of meaning (how truth- functional utterances are produced, how syntactically well-formed phrases are generated), pragmatics stands resolutely on the side of the mechanisms of understanding. The grasp of meaning is only possible because there are regularities governing all transposition of meaning. These regularities are not statistical: the prediction of what the semantic value of an utterance will be is not made on the basis of probability, but on the basis of necessity, which is not falsifiable through counterexamples, deviations, marginalities or exceptions. To grasp meaning is to understand the acceptable. However, the acceptability of a discursive sequence justifies itself on the basis of a schema of reasons which owe nothing to the frequency and probability of phenomena, but everything to certain pragmatic norms which are by no means of an empirical or positive order. If one evaluates the semantic value of a discursive sequence on the basis of the specificity of the mechanisms of its understanding, it is necessary to determine the acceptibility of what one is understanding, as a web of reasons. These reasons have been called principles and maxims (Grice), or conditions (Searle) — one can also call them "constitutive rules". Where does the normativity of this web of reasons that makes the grasping of sense, and thus interlocution, possible, come from? I will linger for a moment on the condition of sincerity (Austin/Searle) and the principle of cooperation (Grice) in order to introduce this epistemological problematics