This dissertation investigates cross-generational linguistic differences in the Canberra Vietnamese bilingual community, with a particular focus on Vietnamese as the heritage language. Specifically, it documents the vernacular and considers key aspects of this data from different theoretical perspectives. Its main contribution is an insight into a rarely studied heritage language variety in a contact community that has never been examined. The dissertation consists of five core chapters, organised into two parts. In the first part (Chapters 2–3), I describe how I documented the vernacular and created the Canberra Vietnamese English Corpus (CanVEC), an original corpus compiled specifically for this study that is also the first to be freely available for research purposes. The corpus consists of over ten hours of spontaneous speech produced by 45 Vietnamese-English bilingual speakers across two generations living in Canberra. In the second part of the study (Chapters 4–6), I put the corpus to use and investigate aspects of the cross-generational differences in Vietnamese as the heritage language in this community. In particular, I first probe the Vietnamese heritage language via its participation in the code-switching discourse (Chapter 4). In doing so, I focus on the applicability of the Matrix Language Framework (MLF) (Myers-Scotton, 1993, 2002) and its associated Matrix Language (ML) Turnover Hypothesis (Myers-Scotton, 1998) to the code-switching data in CanVEC. Since support for this prominent model has mainly come from language pairs that have different clausal word order or vastly different inventories of inflectional morphology, Vietnamese-English as a pair in which both languages are SVO and essentially isolating offers a tantalising testing ground for its application. Results show that the universal claims of this model do not hold so straight-forwardly. CanVEC data challenges several assumptions of the MLF, with the model ultimately only being able to account for around half of the CanVEC code-switching data. I further demonstrate that even when the ML is putatively identifiable and a cross-generational ML ‘turnover’ is quantitatively observed, the predictions do not reflect the direction of structural influence that we see in CanVEC. The MLF approach therefore sheds only limited light on cross-generational language shift and variation in this community. Given that null elements emerge as a distinct area of difficulty in Chapter 4, I take this aspect as the focal point for the next part of the investigation (Chapter 5), where I use the variationist approach (Labov, 1972 et seq.) to explore three cases where null and overt realisation alternates in Vietnamese: subjects, objects, and copulas. In doing so, I move away from the bilingual portion of CanVEC to examine the monolingual heritage Vietnamese subset directly. Results show that Vietnamese null subjects vary significantly across generations, while null objects and copulas remain stable in terms of use. As speakers also overwhelmingly prefer overt forms over null forms (∼70:30) across all the three of the variables of interest, I appeal to the generative interface-oriented approach (Sorace & Filiaci, 2006 et seq.) to next examine the distribution of overt subjects, objects, copulas (Chapter 6). These results converge with what was found for null forms: cross-generational effects were observed for pronominal subjects, but not pronominal objects and copulas. This finding also supports the importance of a distinction drawn in previous works between internal (syntax-semantics) and external (syntax-discourse/pragmatics) interface phenomena, with the latter being seemingly more susceptible to change. Ultimately, this dissertation highlights the empirical and theoretical value of studying rarely considered contact varieties, while deploying an integrated approach that acknowledges the multi-faceted complexity of the contact communities where these varieties are spoken.