Abstract
This paper deals with the making of a corpus of bilingual speech in South Tyrol aimed at exploring language contact patterns in the area. The languages involved are local varieties of Bavarian German, Italian and Trentino, the Italo-Romance dialect of the surrounding area. The paper will briefly discuss the basic
issues concerning the foundation and definition of a bilingual corpus, as opposed to monolingual and to parallel corpora. After a short description of the contact situation in South Tyrol, we will introduce some examples of code mixing extracted from this systematic collection of conversational data and discuss their significance in a theory of bilingual speech.