This thesis has two major aims. The first aim is to answer the research question: is there a universally applicable codeswitching model? The Matrix Language Frame (MLF) model proposed by Myers-Scotton (2002) has been successfully applied to different language pairs, and is claimed to be universally applicable. Muysken (2000), however, argues against the existence of any single codeswitching model with universal applicability, and proposes a typological approach to categorise the patterns of all codeswitching phenomenon. To fulfill the first aim, these two prominent codeswitching models were tested with the Mandarin/Tsou and Mandarin/Southern Min CS data, which were collected from informal conversations by bilingual Mandarin/Tsou and Mandarin/Southern Min speakers located in the Alisan and Tainan regions in Taiwan.
The results revealed difficulties in applying both models to the data. To solve these problems, modified versions of both models were tested. The results of the re-analysis showed that the modified MLF model could be successfully applied to
most of the Mandarin/Tsou and Mandarin/Southern Min data, while the modified typological model was still problematic. Since the original MLF model has been successfully applied to many other language pairs and its theoretical problem has been solved by my modification, I then argue that the revised MLF model proposed in this thesis is a universally applicable codeswitching model. The second aim of this thesis is to predict the future of an endangered language, namely Tsou. To fulfill this aim, further qualitative and quantitative analyses of the results of the re-analysis of the Mandarin/Tsou data by using the revised MLF model were conducted. The Matrix Language (ML) Turnover Hypothesis proposed by Myers-Scotton (1998) was adopted to interpret the results of the analyses.
The results showed that an ML turnover was complete and language shift was in progress in the Tsou language community. Hence, I argue that the survival of Tsou will be problematic.