Abstract
This paper discusses the various challenges encountered in documenting endangered languages spoken in extremely small language communities by speakers who are not always fluent in those languages. On the one hand, some specificities of these languages will be considered, namely their predominantly oral nature and the inherent multilingualism of these communities. On the other hand, some considerations on the nature of the linguistic datum in relation to different elicitation techniques will be proposed. The following reflections are based on a long-lasting fieldwork experience in Walser communities in Italy.