Here, I am going to argue in favour of the corpus-based approach to linguistic typology, since there are many languages for which the texts of a ‘new generation’, i.e. fully glossed and annotated texts with translations and attached audio files, are already available, and even more will appear in the nearest future. On the other hand, grammars and dictionaries of the majority of the so-called endangered languages, if there are any, are still very far from being complete.