International audienceThis study reports on language mixing in two Romani communities, with a century-long presencein Finland and in Greece respectively. A quantitative analysis of free-speech data shows that verbsfrom the contact languages, Finnish and Turkish, are systematically inserted into a dominantRomani speech with their respective Finnish and Turkish tense, mood, aspect, and personmorphology. The insertion in language A of non-integrated single words from language B is atypicalfor classic code-switching and borrowing, but is a well-known mechanism in the creation of mixedlanguages. Unlike mixed languages, however, where no single dominant language can be identified,Romani is the main component in the corpora under study. We suggest that this type of Romanilanguage mixing illustrates an early stage of mixed language formation that did not develop into anindependent mixed language, owing to changes in the sociopolitical settings