Chapter 1 serves as the general introduction to this thesis. Changing from one established language to another within a particular context appears to be challenging. The thesis throughout refers to this phenomenon as the linguistic habit context. This chapter introduces the role of the behavioural sciences and its potential to contribute to increasing the use of lesser-used languages, thus shifting to a more flexible linguistic context. Particularly, how the behavioural sciences can shift the perception that English is the dominant 'language of work' in Wales. One of the core themes of this chapter, and the thesis overall, focusses on the aims of the Cymraeg 2050 vision in terms of increasing the use of Welsh in the workplace. Clearly, the issue of changing language and operating a bilingual workplace is complex. Chapter 1 discusses some of these issues. One of the main aims of Cymraeg 2050 is influencing spoken Welsh in the workplace. This is the focus of the empirical chapters within this thesis (Chapter 2, Chapter 3, and Chapter 4). In addition, the general introduction emphasises a gap in the literature, namely the lack of robust and systematic evaluations of the impact and effectiveness of language interventions on actual language use. These empirical chapters used mixed methods approaches, utilising a number of different methodological traditions in order to attempt to understand these complex issues. Chapter 1 also discusses the widespread use of self-report methods typically used in this area, including their limitations, which bring the strength of previous research into question. A priority for this PhD was piloting and developing observations-based methods, which are more robust than self-reports, in order to inform knowledge regarding the impact and effectiveness of language change approaches and interventions more strongly. Chapter 2 contains two studies. Study 1, Chapter 2 focussed on the development of the Bilingual Dynamic Observational Tool (BilDOT) in order to measure language choice behaviour directly via observations and gather data that are more robust on real-world language choice behaviours within bilingual settings. Study 1, Chapter 2 piloted the BilDOT by observing the language choice behaviours of a group of workers (n=16) in the natural workplace context. The results of Study 1, Chapter 2 demonstrated the BilDOT to be a practical measure for collecting language use data via direct, in-situ observations. As such, Study 2, Chapter 2 used the BilDOT to gather language use data during multiple study phases at a bilingual workplace in order to evaluate the potential impact of a Welsh language mentoring scheme on Welsh use. The results of Study 2, Chapter 2 suggested that the scheme increased the use of Welsh by the participants (n=22). This makes a significant contribution to bridging the current gap of implementing language-driven interventions without adequate evaluation of their impact on language choice behaviours. Overall, the two studies in Chapter 2 demonstrated that the BilDOT was a practical measure for quantifying language choice behaviours. Therefore, the BilDOT allowed the shift away from self-report methods to a more robust data collection method. Chapter 3 took a qualitative approach to evaluating the potential impact of the Welsh language mentoring scheme on Welsh use by interviewing (semi-structured) participants of the scheme (n=25) in order to gather their perceptions of their experience of the scheme. Thematic analysis of the interview transcripts revealed three main themes; (1) changing language is difficult, (2) benefits of scheme, and (3) limitations of scheme. The data suggest that the participants were of the opinion that the scheme was effective in terms of increasing the use of Welsh at work. This is in concordance with the quantitative results of Study 2, Chapter 2. In addition, the data serves as advice in terms of what elements of the scheme were effective in this achievement, and suggest limitations of the scheme. As such, this data can serve as a guide in terms of developing and strengthening the scheme for future implementation. Chapter 4 shifted away from the Welsh language mentoring scheme paradigm of Chapter 2 (study 1 and study 2) and Chapter 3 (study 3) and piloted the ARFer programme (study 4). Integral to the ARFer programme are two behaviour-change principles, namely default setting and public commitment making. ARFer 'enablers' at the department (n=5) committed to use Welsh as the default language choice with their colleagues that could understand Welsh. The use of the BilDOT for gathering language use data was integral to Study 4, Chapter 4 in order to evaluate the potential impact of the ARFer programme on Welsh use (as was the case in Study 2, Chapter 2). However, participants of ARFer (n=22) also completed a questionnaire on a weekly basis in order to self-report their own perceptions of the linguistic dynamic at the department. Thus, two datasets formed the results of Study 4, Chapter 4. Whilst there was inevitable disparity between both datasets, the observations results demonstrated that Welsh use more than doubled during the post-test phase in comparison to the baseline phase. Chapter 5 serves as the general discussion for this thesis. It provides an overview of the thesis, including summarising the empirical chapters that make up its core (Chapter 2, Chapter 3, and Chapter 4). The general discussion also highlights some limitations of the empirical chapters, makes suggestions in terms of future research directions, and suggests some implications of this PhD journey. In addition, the general discussion includes a short reflection on the steps taken in order to minimise participant reactivity, and discusses researcher positionality from the point of view of the author.