Reports on language mixing involving Arabic often qualify that language as resistant to constraints operating on other language pairs. But many fail to situate the purported violations with respect to recipient and donor languages, making it impossible to ascertain whether these are exceptional code-switches or (nonce) borrowings; isolated cases or robust patterns. We address these issues through variationist analysis of Tunisian Arabic/French bilingual discourse. Focusing on conflict sites that reveal which grammar is operative when the other language is accessed, we compare quantitatively the behavior of lone French-origin nouns in Arabic with their counterparts in both donor and recipient languages. Despite a higher order community resistance to morphological inflection of other-language items, results show treatment of French nouns to be consistent with the (variable) grammar of Arabic and different from that of French. Applying the same accountable methodology to the contentious French det+n sequences (“constituent insertions”) shows that most are integrated in the same way as their lone counterparts. These too are treated as (compound) borrowings, largely motivated by the semantic imperative of expressing plurality while eschewing inflection. As borrowings, they do not constitute exceptions to code-switching constraints, confirming that the status of mixed items cannot be determined in isolation; they must be contextualized with respect to the remainder of the bilingual system, including donor, recipient, and other mixed-language elements.