The use of multiple, closely-related languages in childhood is somewhat of a final frontier for several strands of research: child language, second language acquisition, and creole studies. However, the methodologies common in these fields do not easily import to contexts that are typical of remote communities in Australia. In one such community, young Alyawarr children go about their daily lives mainly speaking Alyawarr English, a new contact language. At school they use Standard Australian English as a second language. Much of what they encounter in SAE seems familiar and much is obviously different. In between, there are subtle differences that are possibly harder to detect, parse and maintain. This study takes a corpus of 50+ hours of naturalistic video recordings of six focus children, and creates two maximally contrastive, contextually defined data sets. Shared, variable features are analysed using the Comparative Variationist method. This study breaks new ground in both methodological terms, with the application of variationist modelling to child bi-varietal language repertoires, and in advancing our understanding of the vectors of repertoire management in the complex ecologies of the region.