Documenting multilingual language practices and the erasure of language boundaries

International audience Research on linguistic variation, language contact or codeswitching mostly assumes bounded languages or repertoires. Most studies presuppose the contact between stable 'communities' and the identification of specific languages in bilingual (sometimes plurilingual) corpora. Similarly, language annotation in corpus linguistics is based on the principle of univocity of items belonging to specific languages. On the other hand, in an attempt to avoid referring to bounded languages, a lot of terms have been coined in sociolinguistics or anthropological linguistics-as transidiomatic practices, heteroglossia, polylanguaging, translanguaging, languaging etc. Heterogeneous and hard to classify, multilingual language practices are due to, and at the same time reveal linguistic diversity and the multiplicity of self-identification and positionnings linked to processes of homogeneization and differentiation.In this talk, I will address the notion of language boundaries, constructed both by linguists and by language users and will consider variation and heterogeneity as linguistic resources for speakers in their everyday multilingual language practices. It needs first a “shift in focus from linguistic systems toward language users [... and their] repertoires drawn from lived experiences that may disrupt presumed connections between language, community, and spaces” (Hall & Nilep, 2015: 615).Second, it needs also a solid methodology to reveal the heterogeneity of language practices through the annotation of corpora. The annotation method I draw on reveals the heterogeneity of language practices and at the same time shows how languages or varieties may overlap making it irrelevant to draw arbitrary lines and boundaries between linguistic resources. This has of course an impact on the way we – as linguist – consider the attribution of language (or variety) labels to linguistic forms.I will show how ambivalence or a play on boundaries is a common characteristic of communication in highly multilingual contexts. Language users make use of all kinds of linguistic resources in order to communicate and, when they share more or less the same language(s) background (ie in endolingual settings (Lüdi 1987), they sometimes choose ambivalent elements, attributable to various languages or varieties. In doing so, they choose, in a way, not to perform language boundaries but to ‘float’ instead in mid-water. It is particularly obvious in postcolonial contexts (involving for example close language varieties or a Creole and its lexifier in decreolization contexts). The use of bivalent elements may be politicized and controversial (Woolard 1998) or may represent a linguistic resource that is strategically marshaled by language users in their everyday interaction to position themselves, linked to processes of homogeneization and differentiation