In this paper, we examine the construction of observables, in quantitative corpus linguistics, for the description of linguistic genres and discourses and for textual typology. We propose a hypothesis according to which texts may be characterized through correlations between levels of description, rather that using one isolated level of description at a time. The systemic relations between levels of description allows us to use categories of textual linguistics in quantitative analysis. New contextualities have to be considered: contextuality as interaction between levels of description, and contextuality as interaction between genres, discourses, and idiolects in each text. Coupling textual linguistics and quantitative corpus linguistics raises the issue of the relation between textual typology and linguistic typology.