The reconstruction of the diachronic origin of a grammatical formative does not boil down to an exercise in etymology. It does not suffice to name some plausible phonological and semantic changes which led from the original to the present form. It must also be shown how the formative ended up in its present construction with its paradigmatic and syntagmatic relations.

In the most straightforward case, the sign in question originates in a certain textual or syntactic construction where it is motivated by the conventions and rules of that level; and with the grammaticalization of that construction, the sign in question becomes a grammatical formative whose paradigmatic and syntagmatic relations – especially, its syntagmatic position and combinatory potential – are just a direct continuation of its original function, formalized along the parameters of grammaticalization. For instance, saying that the Romance adverbializer -mente stems from the Vulgar Latin noun mente ‘mind’ is insufficient. What one must show is how, in Vulgar Latin, a complex nominal constituted by mente as its head noun and modified by an adjective as its attribute acted as an adverbial in the ablative to modify some predicate, as when tranquilla mente ‘with a calm mind’ becomes Italian tranquillamente ‘calmly’.

While such a course of things is the default hypothesis in many cases, it is not always what actually happened. To postulate it absolutely may be dubbbed the ‘marker/structure fallacy’. The marker/structure fallacy consists in the automatic assumption “that the structure in which a particular innovative grammatical element is found developed out of the structure in which that grammatical element originated” (Harris & Campbell 1995:284). For instance, English relative clauses introduced by who do not go back to interrogative clauses just because who is originally an interrogative pronoun. In fact, who replaced the older subordinator þe in an existent relative construction.

Thus, to repeat: Trying whether the origin of a given formative can be reconstructed along the lines of a grammaticalization path is good methodological practice, and certainly better than a mere etymology; but it is not a guarantee of the way to truth.