With a view to analyzing the way media deal with specific "discourse moments" (such as the debates on GMO's), two stages are drawn up. First, a survey is made of the descriptive categories that prove necessary to extract "observables", which are located along texts and documents, and also to set up the "sub-corpora" they bring up. But in the purpose of contextuualizing these data, we need to make use of the operating notion of dialogism, with its various forms of actualization. Thus, in a second stage, it is showm how the first exploratory corpus can be expanded to interdiscourses, which come up as reminders of previous discourses and facts: this interdiscurisve memory bank, built up by the media, takes us back to the history of the relations between science, nature and society.