A CORPUS-BASED ANALYSIS OF METADISCOURSE MARKERS IN THE DISCUSSION SECTION OF RESEARCH ARTICLES: A COMPARATIVE STUDY OF SOFT SCIENCES AND HARD SCIENCES

Abstract

This study investigated the use of metadiscourse markers—key rhetorical devices that structure academic writing and facilitate writer–reader interaction—in research articles. While previous research has acknowledged disciplinary variation in metadiscourse use, this study specifically examined how L2 writers employ these linguistic resources in the discussion sections of research articles across the soft and hard sciences. Using a corpus-based approach, the analysis explored the frequency, functional roles, and disciplinary appropriateness of metadiscourse markers. The corpus consisted of 100 discussion sections from reputable journals indexed in Scopus and ScienceDirect, representing both disciplinary groups. Hyland’s (2005) interpersonal metadiscourse framework and AntConc software were used to analyze both interactive markers (e.g., transitions) and interactional markers (e.g., hedges, boosters) through quantitative (normalized frequencies) and qualitative (concordance) methods. The results indicated a significant statistical difference in metadiscourse density, with soft sciences showing higher usage (M = 117.2 per 1,000 words) than hard sciences (M = 86.5). Soft sciences favored interpersonal markers, particularly boosters (e.g., we argue), whereas hard sciences relied more on interactive transitions and hedges (e.g., may suggest). Findings also revealed strong adherence to disciplinary conventions, with subtle L1-related and cultural influences. The study proposes a genre-based instructional approach tailored to disciplinary needs and recommends future research on sub-disciplinary variation and longitudinal development of L2 writers’ metadiscourse use.