Lexical Bundles across New Engineering Disciplines: A Corpus-Based Comparison of Research Articles for Teaching Academic Writing

Abstract

This study investigates disciplinary variation in four-word lexical bundles in research articles from three representative New Engineering Disciplines: artificial intelligence, biomedicine, and robotics. Three comparable, recent corpora were compiled from internationally indexed journals, consisting of full-length research articles. Lexical bundles were extracted using AntConc with frequency and dispersion thresholds selected to balance representativeness and analytical manageability. The identified bundles were then categorized structurally using Biber et al.’s (1999) model and functionally using Hyland’s (2008a) framework, with minor extensions introduced to accommodate recurrent patterns not captured by the original taxonomies. Results reveal both cross-disciplinary convergence and discipline-specific preference. Structurally, the three corpora display a profile typical of hard-knowledge research writing, with a substantial presence of verb phrase (VP)-based and other clausal patterns alongside noun phrase (NP)-based and prepositional phrase (PP)-based bundles. Biomedicine shows a comparatively stronger reliance on PP-based framing resources, consistent with the need to specify conditions and constraints, whereas artificial intelligence and robotics exhibit relatively stronger VP/clausal patterning associated with reporting and evaluation. Functionally, research-oriented and text-oriented bundles dominate across disciplines, while participant-oriented bundles are comparatively limited, particularly in biomedicine. The findings underscore the value of discipline-sensitive, corpus-informed EAP instruction for New Engineering students.