It can be seen in the massive use made by humanities of metaphor and borrowing to the exact sciences as best as a desire to object, at worst an impressive strategy, sometimes confined to scientific fraud. Without wishing to take part in this useful debate, we propose to study the ways in which the linguist Gustave Guillaume (1883-1960) regularly used mathematics, biology and physical sciences throughout the process of theorising her psychomecanic language, concepts or words (with only their meaning). The study is part of a research on the linguistic analysis of insoved texts of digital philology. Because it makes the double assumption that a theory is a text and that theorisation is part of the construction of meaning, this research, which we call digital epistemology, uses its tools for analysis and description using textual semantic (F. Rastier) and corpus language investigation techniques. From a methodological point of view, we focus here on the return of the intertext (or simulated intertext). The study corpus includes the full text of 30 conferences delivered between 1958 and 1960. The whole is about 100 000 words.