One difficulty encountered in using theoretical approaches, which aim to formalize natural languages, is that the linearity of phonic sound matter is misleading, since it hides the non-linearity of a large number of syntactical and semantic processes that contribute to the specific structuring of each language. Therefore, it is well-known that the actual idea of « transformation », characteristic of Chomsky’s generative model, came about in order to account for the linear discontinuity of certain syntactical patterns that required the use of rules of movement and erasing for their formalization. Yet, the categorial approach, which in the twentieth century was the main formal grammar trend able to compete with the transformational model, was opted for – contrary to Chomsky’s choice – to elaborate a syntax in accordance with the linearity of languages. The aim of this communication is to examine how the categorial approach – while respecting linearization – succeeds in analyzing syntactical structures which transgress the linearity by using processes such as embedding or interlacing. Thus, we’ll begin by pointing out the methodological constraints imposed by the categorial approach in order to ensure that calculations related to grammatical correctness and the interpretation of utterances obey the linearity principle. Then, we will examine two types of examples which do not comply with linearity and which present an increased degree of complexity. First, we will analyze how Lambek’s pre-group categorial grammar succeeded in linearly formalizing the problem of “nested dependencies” in German subordinates, such as “…, ob ich dich sehen will”. Next, we will study the categorial treatment proposed by M. Moortgat to solve the more difficult problem of “crossed dependencies” of verb groups in Dutch clauses (i.e. … omdat Alice de Soepschildpad will plagen). The comparison of two very distinct categorial strategies will lead us to the conclusion that respecting linearization cannot be obtained simply through local modifications because in all theories it is necessary to make in-depth syntactical changes.