Abstract Historical metalinguistic discourse is known to often prescribe linguistic variants that are not very frequent in actual language use, and to proscribe frequent variants. Infrequent variants that are promoted through prescription can be innovations, but they can also be conservative forms that have already largely vanished from the spoken language and are now also disappearing in writing. An extreme case in point is the genitive case in Dutch. This has been in decline in usage from at least the thirteenth century onwards, gradually giving way to analytical alternatives such as prepositional phrases. In the grammatical tradition, however, a preference for the genitive case was maintained for centuries. When ‘standard’ Dutch is officially codified in 1805 in the context of a national language policy, the genitive case is again strongly preferred, still aiming to ‘revive’ the synthetic forms. The striking discrepancy between metalinguistic discourse on the one hand, and developments in language use on the other, make the genitive case in Dutch an interesting case for historical sociolinguistics. In this paper, we tackle various issues raised by the research literature, such as the importance of genre differences as well as variation within particular genres, through a detailed corpus-based analysis of the influence of prescription on language practices in eighteenth- and nineteenth-century Dutch.