After classifying the sociology of language in three paradigms and showing the shortcomings of the terra «linguistic attitude», a comparison is made between linguistic substitution processes and butterfly catastrophes. The similarities found are useful because 1) they provide us with a global explanatory model based on a neutral terminology, 2) they show us the number of variables that intervene in the process; 3) they allow us to describe both individual and group behaviour with a higher degree of penetration, and at the same time, in a way easier to understand; 4) finally, they provide us with a model that could not be conceived of so far without resorting to catastrophe theory, which opens up more consistent prospects than those analyzed in the introduction. In short, by applying catastrophe theory to the sociology of language we would litre to come closer to being able to consider what Michel Foucault termed the scientificity borderline, that is, the moment when discourse reflection is made, not according to empirical rules, but to a basic formal system.