This paper ask the question of how the writing of knowledge shapes its objects. Its perspective rejects the notion of a language as a mirror or as the reflection of a reality existing elsewhere, and thus of a scientific writing relating discoveries of autonomous facts. On the contrary, it enquires how scientific discourse defines, constructs and achieves the observed facts, as well as their naturalisation and reification as facts belonging to the "reality" of the world. Geographical representations, just like other scientific discourses, constitute the symbolic mediations which furnish the schemes of intelligibility for describing, interpreting and explaining the world. The images which they produce derive from the construction of an intersubjectivity, from the bringing together of scientific paradigms, from the inscription in networks, from the recourse to conventions of writing which are more or less recognised by the Academy - and not from any correspondence between the words and the outside world. In this article, we will develop the theoretical consequences of this position; we will also propose analytic tools for an empirical approach considering both what is at stake in geographical contemporary debates and what a linguistic reflection can contribute to.