<p>This thesis investigates metadiscourse in master’s theses and the relationship between metadiscourse frequencies and quality of thesis writing. Metadiscourse has been a major research focus in various genres and contexts, but only a small proportion of this work has compared metadiscourse in postgraduate writing across educational contexts and disciplines. While previous studies of metadiscourse have reported a positive correlation between metadiscourse frequencies and writing quality, all of these studies focused on undergraduate writing. Little is known about the relationship between metadiscourse frequencies and quality of thesis writing. This thesis includes two main studies to address the gaps in literature. Study 1 examined use of metadiscourse (i.e., frequencies, types, and functions) in master’s thesis discussion and conclusion chapters written in English by New Zealand and Thai postgraduates in the disciplines of English language teaching and business administration. Four subcorpora with a total of 116 thesis samples were compiled: 26 New Zealand students’ theses in English language teaching (NZ-ELT), 30 New Zealand students’ theses in business administration (NZ-BA), 30 Thai students’ theses in English language teaching (TH-ELT), and 30 Thai students’ theses in business administration (TH-BA). Hyland’s (2005) metadiscourse taxonomy was adopted for this study. Study 2 explored the relationship between metadiscourse frequencies and quality of thesis writing. Forty eight theses (twelve theses with highest and lowest frequencies of metadiscourse markers in each of the four subcorpora in Study 1) were selected for this study. Semi-structured interviews were conducted with twenty four disciplinary supervisors in New Zealand and Thailand in order to investigate supervisors’ attitudes towards and expectations of good thesis writing in their disciplines. This information was used to design a rating scale specifically for thesis quality assessment. Two New Zealand and Thai raters in English language teaching and business administration, four raters altogether, rated twenty four discussion and conclusion chapters in their own disciplines. The findings of Study 1 reveal a higher frequency of metadiscourse in New Zealand theses than Thai theses. While both New Zealand and Thai students use more textual metadiscourse than interpersonal metadiscourse, New Zealand students show a greater reliance on the use of interpersonal metadiscourse (all interpersonal subcategories, except for boosters) than Thai students. By contrast, Thai students show a greater reliance on the use of textual metadiscourse (especially transition markers and frame markers) than New Zealand students. With regard to disciplinary variation, English language teaching students use more metadiscourse than business administration students, in both textual and interpersonal metadiscourse categories. Transition markers and hedges are the most prominent metadiscourse features, contributing the preponderance of textual and interpersonal metadiscourse in this study. Despite palpable differences in frequencies, the analysis of individual types across the four subcorpora reveals similarities of New Zealand and Thai students in the two disciplines. They use similar markers, rely heavily on a small cluster of high frequency markers, and make scarce use of lower frequency ones in all subcategories. The functional analysis indicates that there are six subcategories whose functions contribute to differences between New Zealand and Thai theses, namely transition markers, frame markers, evidentials, attitude markers, engagement markers, and self-mentions. The findings of Study 2 reveal a positive correlation between metadiscourse frequencies and thesis quality scores in both English language teaching and business administration disciplines. However, a major difference between the two disciplines is that in business administration, the frequency of textual metadiscourse is more highly related to the quality scores when compared to interpersonal metadiscourse. In English language teaching, the frequency of interpersonal metadiscourse is more closely related to the quality scores. In the comparison of quality scores between high and low frequency groups, a statistically significant difference is found in business administration, but not in English language teaching. Insights gained from this study are that (1) business administration raters are likely to focus more on textual features which directly affect readers’ comprehension, while English language teaching raters seem to have more expectations towards interactional features (e.g., explicit expression of students’ attitudes towards their own research propositions), (2) not all metadiscourse subcategories affect thesis quality scores, and (3) apart from frequencies, factors such as appropriate use of a wide variety of markers in different subcategories may contribute to better quality scores. Based on these findings, this thesis also provides theoretical, methodological, and pedagogical implications, laying out a framework for postgraduate writing instructors in developing English for Postgraduate Academic Writing lessons and materials based on actual language use and expectations of members in specific disciplinary communities and educational contexts in order to improve postgraduate writing quality.</p>