The aim of this paper is to explain and evaluate Rorty’s thesis of the contingency of language, by examining the validity of the Davidsonian proposal of passing theories of meaning, upon which the thesis rests. The impossibility of the existence of different vocabularies irreducible to a single one, or the impossibility of a definitive meta-lexicon (contingency thesis), presupposes the validity of the Davidsonian approach of passing theories, which attempts to dissolve the notion of language as a defined a priori object of study that could stand a theory. Now such Davidsonian approach to language and communication faces a difficulty: the very possibility of understanding the language deviations posed by malapropisms presupposes knowledge of strict semantic rules. Such assumption can be interpreted as the need to recognize that there are transcendental conditions of communication, as Habermas argues, which implies that Rortian contextualism involved in the contingency thesis, and the thesis itself, must be false. I will show that Rorty can counter-attack Habermas’ transcendentalism by a critique of the transcendental model of argumentation, which attempts to show that it is meaningless to speak of transcendental conditions that we can apply and learn, which supposedly constitute the mechanism to overcome the contextualism involved in the thesis of the contingency of language.