the possibility for a name to offer a taxonomic reading (X is an Y/subclass of Y) and a blurred reading (X resembles Y without being a subtype of Y) can constitute a discriminatory typological parametre? This is the guiding thread of our query. This will involve testing different types of N, from the most standard to the most atypical, and understanding the semantic and categorical mecanisms that allow or block these two readings. Initially, which is the subject of this article, the examination relates to three types of N (accounting N, solid N and N colour) and makes it possible to measure a number of criteria which are clearly distinguishable between approximate and non-approximate categories.