Although observation of usage has always been an essential empirical parameter in the development of linguistic theories, its consideration varied, to very different degrees, during the second half of the twentieth century. From the 80’s, in opposition to the very poor, reductive and sometimes distorted idea of the notion of usage brought into play in Chomsky’s syntactic models, new approaches emerged that chose to consider usage not as the updating of a system, but as the cause behind the emergence of and change in linguistic forms. The Chomskian postulate of the autonomy of grammar found itself deeply questioned, instead favoring approaches that considered grammar to be affected, and even shaped, by the use of natural languages. The new models which resulted – whether functionalist or cognitive – became depositaries of dynamic syntax approaches and were constructed according to the frequency of exposure to language data, the categorization of exemplar forms and the influence of several processes such as the schematicity of linguistic forms, the accommodation of unconventional data, or analogy. The epistemological primacy thus given to usage – understood in its double meaning of “speech act” and “habit” – resulted in the revolutionizing of the overall architecture of linguistic theories, the revising of the concepts and dichotomies used to describe natural languages and also the modifying of the idea of the nature to be attributed to language. The purpose of this communication is to lay out the theoretical changes induced by usage-based approaches to language in order to bring to light the causes of the current vitality of this movement. Thus, I will begin by highlighting the challenge to the canonical order (grammar/lexicon - semantic - pragmatic) induced by these approaches as well as the change in definition that these different fields have undergone. Then I will examine the questioning of commonly used concepts (the notion of unit of language, of rule, etc ...) and show why the dichotomies local/global, language/speech and competence/performance no longer find a foundation in usage-based approaches. Finally, I will explain why, in this kind of approach, natural languages are described as complex adaptive systems. When usage takes the lion's share in the construction of linguistic theories, language can then be conceived as a natural and social system in continual evolution.