Given the traditional connection between the truth of analytic sentences and their sense, one can ask what happens to analyticity in Dummett's model for a theory of meaning. Dummett claims, against Quine, the legitimacy and necessity of the talk of sense: sense is in fact, according to his model, the object of one part of a theory of meaning (theory of sense and theory of reference form its first part, the second part being the theory of force), for no description of how language works can do without an account of what a speaker knows when he speaks competently. Now Dummett's defence of the notion of sense can be understood as a defence, at the same time, of the notion of analyticity: the traditional connection between sense and analyticity is so kept, just like Quine did, the other way round, in criticizing those notions. The traditional definition of analytic sentences as the ones which are true in virtue of their sense turns out nevertheless to be misleading, first of all because Quine's criticism does score a point: some analytic sentences are susceptible to turn out not to be true, if nor to be actually false. Their truth is contingent, for they presuppose other sentences that can possibly be false; now, presupposition being not identical to implication, analyticity does warrant against falsehood, but it does not guarantee truth. Moreover, the link between analyticity and sense fits only one class of intuitively analytic sentences; now, while letting programmatically aside the 'extensionally' analytic sentences (namely those whose truth does not follow from the mere sense of their components, demanding instead logical and/or mathematical axioms having substantial implications), this paper proposes, within a Dumettian framework, a theory of analyticity for the sentences which are true in virtue of their sense and for the sentences of the theory of sense themselves, where the analyticity of the former is strictly tied to that of the latter. A sentence describing the use of an expression is analytic because it makes it explicit how such an expression is used, while some sentences (viz. 'no bachelor is married') are analytic just because of this use, although they do not themselves describe it. Since descriptions are in general informative, for each trivial analytic sentence one can find a nontrivial one describing the use of one expression occurring in it ('bachelor'); and therefore the use of at least one nontrivial and nonanalytic sentence ('x is bachelor'), i.e. the canonical procedure for establishing it (by verification or by resistance to falsification). Now, in describing the use of other sentences, the sentences of a theory of sense do quite a peculiar metalinguistic performance: they strive to say what the latter simply show, although without mentioning, even less reifying, their sense. [ABSTRACT FROM AUTHOR]