Corpus tools have long been recognised as useful for teaching academic writing, particularly in multilingual and cross-disciplinary contexts. However, most current corpus tools do not directly integrate the genre-based pedagogical approach advocated by Swales and Feak, which emphasises exploring how communicative purposes shape generic structure and phraseology, while considering cross-cultural differences. In response to this gap, this paper presents the EXEMPRAES Corpus tool, a bilingual, genre-sensitive resource that integrates a communicative-function-annotated comparable corpus with a web-based interface designed to operationalise this pedagogical orientation for cross-cultural academic writing instruction. The paper explains how Swalesian principles have informed the tool's design at multiple levels, from corpus compilation, segmentation, and annotation to interface features that support sequenced, discovery-based tasks moving from audience and purpose to organisation, style, flow, and presentation. Drawing on expertvalidated annotations of discussion and/or conclusion sections in ten pairs of socialscience research articles in English and Spanish, we illustrate how the EXEMPRAES Corpus tool enables exploration of cross-cultural generic and phraseological variation at the level of specific communicative functions, thereby supporting genre-based pedagogy in multilingual academic contexts in ways not typically available in standard corpus interfaces. This proof-of-concept implementation demonstrates how this approach, enriched by insights from research on intercultural rhetorical and disciplinary variation, can be operationalised in a user-friendly interface for pedagogical purposes. We also suggest that this design could serve as a model for future applications with other language pairs, illustrating how Swales's ideas can continue to shape future work. (c) 2026 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Teachers and students are increasingly exploring how Generative AI (GenAI) tools can promote learning and increase efficiency in carrying out common communicative tasks in English as a second or foreign language. One important yet underexplored domain is student-faculty email communication, a genre where learners are expected to manage linguistic fluency, pragmatic appropriateness, and interpersonal decorum associated with hierarchical student-faculty relationships. The present study investigates how Chinese learners of English use DeepSeek to compose their emails and examines faculty members’ perceptions of GenAI-assisted emails in comparison to students’ self-written ones. Students (N = 53) from three English-medium universities in China completed two email-writing tasks, wherein their prompts fed to DeepSeek were recorded and categorized using Flower and Hayes’ (1981) framework. Results showed that students rarely engaged GenAI for planning, but adopted a generate-then-refine strategy. This quantitative process data was triangulated with survey and interview data on student and faculty perceptions. Students viewed GenAI as easy to use and valuable for demonstrating politeness and clarity, although they also noted limitations such as overly formal or verbose phrasing. Faculty members rated GenAI-assisted emails as more polite and professional but not necessarily clearer or more persuasive, and some raised concerns that overly polished language might reduce the sense of personal voice. The findings provide new insights into how GenAI is reshaping the writing process for academic email communication, while also revealing a tension between student and faculty perceptions of GenAI-assisted writing.
This study investigates the perceptions of second language (L2) learners regarding artificial intelligence (AI)-generated feedback on speaking skills within an English for Specific Purposes (ESP) context. While AI-generated feedback has shown promise in writing instruction, its impact on real-time L2 speaking performance remains underexplored. The research was conducted at a private university in Istanbul, Turkey, utilizing a sample of 80 students who enrolled in mandatory ESP courses for Logistics Management (n = 39) and Civil Aviation and Cabin Services (n = 41) departments. A mixed-methods research design was employed, incorporating an intervention period of four weeks. The experimental group received automated feedback via ChatGPT derived from Microsoft Teams transcripts of classroom discussions, while the control group received traditional synchronous instructor feedback. Quantitative data were gathered using Feedback Orientation Scale (FOS) and a modified version of FOS to assess perceptions of utility, accountability, social awareness, and self-efficacy. Qualitative insights were obtained through semi-structured interviews and inductive thematic analysis. Findings indicate that while learners recognize the diagnostic value and speed of AI, they significantly prefer human feedback for its emotional resonance, perceived trustworthiness, and contextual nuance. These results suggest a need for hybrid feedback models that balance AI scalability with human empathy in specialized language instruction. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Job application letters (JALs) are a key employment-related genre through which applicants present their qualifications, experience, and suitability for a specific role. Yet some graduates may find it challenging to write letters that are sufficiently specific, rolerelevant, and persuasive. This raises an important concern for ESP pedagogy about how students can be supported to write application letters that are more employer-oriented. This study investigates the linguistic impressiveness of JALs written by ESL graduating students at a Malaysian research university. In this study, linguistic impressiveness refers to how effectively a JAL uses genre structure, alignment with job-advertisement requirements, and persuasive self-presentation to make the applicant's professional suitability visible. Drawing on Fairclough's three-dimensional model of discourse and Upton and Connor's move structure framework, the study analysed 50 JALs in terms of move structure, lexical alignment with job advertisements, and persuasive self-presentation. The findings show that students generally included the main JAL moves, especially applying for the position and providing supporting information, but were less consistent in realising closing moves such as expressing appreciation, offering further information, and referring to the attached resume. Many letters also showed only partial alignment with job-advertisement requirements and a slight tendency toward descriptive rather than persuasive self-presentation. The paper concludes by considering how ESP instruction can better support students' development of persuasive and employer-oriented job application writing. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
This discussion note revisits John Swales' epistemology of genre to examine how disciplinary expertise, authority, and belonging are recognized in and through language, often without being explicitly stated. Swales' work has repeatedly pointed to dimensions of specialized discourse that remain tacit & horbar;what is presupposed, backgrounded, or left unsaid. Building on these insights, the paper argues that genre analysis, though highly effective at describing stabilized textual patterns, encounters an analytical limitation when accounting for how readers recognize competence and legitimacy from minimal linguistic cues. Drawing on diachronic and ethnographic research, the paper introduces an indexical genre perspective that extends Swalesian genre analysis by theorizing how meaning, context, and expertise are interpreted through shared inferential frameworks rather than fully encoded in text. Indexicality is used to account for how patterned linguistic cues & horbar;such as omission, compression, and routinized formulation & horbar;activate shared interpretive frameworks through which disciplinary insiders evaluate stance, credibility, and alignment. The paper outlines an approach to indexical genre analysis that links textual cues to sociohistorical disciplinary practices and examines how their interpretation can be traced and empirically tested through reader uptake. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
In second language (L2) academic writing, first-person pronouns are widely recognized for their role in constructing writer identity, yet how this identity work is accomplished through interaction, a defining feature of academic discourse, remains underexplored. In light of this, the present study examined how first-person pronouns function as interactional resources for constructing writer identity in a corpus of argumentative essays written by Chinese students in an English as a foreign language (EFL) context. Drawing on frameworks of writer identity roles and Hyland's (2005) interactional model of stance and engagement, the corpus-based analysis identified a set of fine-grained interactional functions that extend the broad stance-engagement distinction and demonstrated how these functions contribute to the construction of distinct authorial identities. Moreover, Chi-square tests of independence further revealed systematic associations between pronoun forms and the interactional functions they realized. The study contributes to the literature on L2 argumentative writing by integrating interactional functions and authorial identities and demonstrating the value of an interactional perspective on self-mention for understanding how writer roles are enacted. Pedagogically, the findings provide a framework to support students in making function-aware pronoun choices and recognizing how such choices shape writer identity projection. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
With the emergence of LLMs, a growing body of research is delving into the linguistic differences between AI-generated academic written texts and human author-produced ones. However, little empirical research has been conducted to uncover their metapragmatic differences. To bridge the gap, this study investigates the metapragmatic acts performed in research article (RA) abstracts generated by ChatGPT 4o, in comparison to those produced by human authors. It reveals that (1) ChatGPT 4o performs a significantly lower frequency of ideational metapragmatic acts; (2) ChatGPT 4o and human authors perform the same general types of metapragmatic acts, except for subtypes of providing an evidential and signaling mutual knowledge; (3) ChatGPT 4o uses much less diverse linguistic devices when performing metapragmatic acts but overuses certain devices like 'such as', 'these', and 'em dash'. The study contributes to the optimization of LLMgenerated academic written texts, offers a pragmatically informed lens for researchers examining RA abstracts, and provides EAP/ESP instructors with insights into cultivating students' AI literacy, particularly their metapragmatic competence in evaluating and refining LLM-generated academic writing. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Environmental, social, and governance (ESG) reports function not merely as a company's non-financial "transcript" but as a strategic communication tool that influences investor perceptions and fosters stakeholder trust. To uncover the linguistic features of ESG reports, this study employs a complete multidimensional analysis (MD) to analyze 269 ESG reports published by Fortune Global 500 companies. Six functional dimensions are identified: (1) elaborated discourse versus informational density; (2) abstract and formal style; (3) descriptive style; (4) statement of persuasion versus expression of stance; (5) interactive versus non-interactive discourse; and (6) narrative versus integrative discourse. Dimension 1, the most salient co-occurring pattern, shows no significant differences across firms with different performance levels, indicating a genre-defining feature driven by institutional pressures. By contrast, significant differences emerge in dimensions 2, 3, and 5. Reports produced by high-performing firms tend to be less abstract and formal, more descriptive, and more reader-accessible than those of lowperforming firms. This study advances research on corporate discourse by extending the application of MD to the ESG reporting domain. The findings offer empirically grounded guidance for improving ESG communicative effectiveness and provide implications for business English teaching and research. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
This study examined how a conversational agent, customized through advanced prompting in OpenAI's GPT Builder, can be used for speaking assessment in English for academic purposes (EAP). The system was developed around four key dimensions: measurement definition, system design, performance evaluation, and learner experience. First-year English majors from two large universities participated in assessments where they could choose topics, control response time, and request clarification. Using many-facet Rasch measurement and linear mixed-effects models, we found that the GPT system exhibited a mid-range level of scoring severity among raters overall, while mean score differences were not statistically significant. Three out of eight scoring categories showed statistically significant differences between GPT and human ratings. Despite these differences, GPT scoring demonstrated acceptable consistency and reliability while exhibiting typical rater tendencies such as restrained use of extreme scores. Students requesting clarification tended to receive lower scores, while those taking longer speaking durations tended to perform better. Students' perceptions about GPT did not significantly influence scores, which suggests that the test outcomes were not influenced by factors unrelated to the measured speaking ability. Overall, findings indicate that GPT-powered systems may provide a useful and scalable option for formative speaking assessment, especially in classroom settings prioritizing student choice and flexibility.
Since its conceptualization by Swales (1981), Genre Analysis has become one of the most influential and widely adopted frameworks for examining the complexities of academic research writing, leading to extensive application in pedagogical contexts (Swales, 1990). It has inspired significant developments not only in the field of discourse and genre analysis as a key contributor to English for Specific Purposes, but more broadly in applied linguistics by expanding its scope to cover a number of other professional and disciplinary genres. I have been closely associated with the development of Genre Analysis since its inception in 1981 and have explored its applications beyond academic research writing. In this paper I would like to offer a reflective account of some of the insightful extensions of genre theory, especially the way it has evolved from its focus on textualisation in academic research settings to contextualization in professional, institutional, and other contexts. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
John Swales' work in genre theory and pedagogy includes pioneering theoretical and pedagogical texts that have become a part of the fabric of genre studies and have greatly informed how scholars and teachers understand genre. This paper explores Swales' writing as textual artifacts that themselves can provide insights into genre. In examining characteristic features of John's texts, I consider what his writing can teach us about several aspects of genre, such as identity as transcending genre, how individuals can shape discourse communities, storytelling as a rhetorical strategy, audience awareness as embodied experience, and the critical human dimension of genres as social practices. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Bias has long been viewed as a negative element in language testing, defined as any systematic construct-irrelevant variance that threatens validity and fairness. However, this traditional interpretation becomes problematic in English for Specific Purposes (ESP) assessment, where performance necessarily depends on domain-specific knowledge and professional reasoning. This paper redefines bias as a construct-relevant differentiation & horbar;a legitimate reflection of the professional and contextual specificity embedded in ESP constructs. The concept of positive bias is proposed to describe systematic variations that arise from the authentic alignment between language performance and domain knowledge, rather than from unfair advantage. The study develops a multi-level framework for the identification, measurement, and validation of positive bias, combining theoretical, operational, and empirical evidence. It further elaborates on the operationalization of positive bias through task design, rating criteria, and validation procedures, and discusses how Differential Item Functioning can distinguish construct-relevant from construct-irrelevant variance. This paper argues that bias in ESP assessment is not a defect to be eliminated but a feature to be calibrated. The proposed framework transforms bias from a marker of unfairness into an indicator of authenticity and construct validity, offering new theoretical and methodological directions for ESP assessment research. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
The field of English for Research Publication Purposes (ERPP) has developed in response to the globalisation of academic publishing and the dominance of English as its lingua franca. Although initially framed within English for Academic Purposes (EAP), ERPP has since emerged as a distinct area of inquiry. It increasingly meets Krishnan's (2009) criteria for disciplinary status, with its own research focus, theoretical frameworks, methodologies, metalanguage, and institutional presence. The field is grounded in foundational studies that have shaped its conceptual and empirical trajectory. These include Swales's theorisation of genre and discourse community, Canagarajah's analysis of structural barriers in global publishing, Lillis and Curry's work on literacy brokering, and Hyland's research on disciplinary identity and stance. Collectively, these contributions highlight the field's commitment to both critical inquiry and pedagogical practice. The paper argues that ERPP has matured into a vital domain of applied linguistic research, one that brings together questions of language, access, and scholarly participation. In doing so, it reflects and extends the legacy of John Swales, whose work has profoundly influenced the field's development and its ongoing engagement with global academic communication. (c) 2026 Published by Elsevier Ltd.
The conclusion chapter of a doctoral thesis is important to the thesis's overall reception and potential impact. However, the linguistic resources used to highlight research significance in conclusion chapters remain underexplored. This study employed SciBERT, a language model pre-trained on scientific texts, to computationally analyse adjectives of importance in 150 doctoral theses across the disciplines of applied linguistics, physics, and psychology. Using UMAP dimensionality reduction, k-means clustering, and permutation tests, this study examined the overall distribution, semantic clustering, disciplinary variation, and syntactic functions of these adjectives. Results revealed that important and significant dominated across disciplines, with a two-cluster (k = 2) semantic structure emerging as optimal. Eleven adjectives showed statistically significant disciplinary differences, with physics theses using most of them less frequently. Syntactic analysis revealed a dominant use of attributive functions. The findings demonstrate how doctoral students strategically deploy adjectives of importance in their thesis conclusions and offer insights for discipline-specific instruction in thesis writing. The computational analysis also extends methodological applications in academic writing research. (c) 2026 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
This study proposes a novel methodological framework, Lexical Multidimensional Analysis of Key Keywords (LMDA-KK), to operationalize the concept of 'variability within stability' in diachronic discourse research. Using China Daily's international reporting on China (2017 -2024) strictly as a proof-of-concept, the study investigates the stable, trans-topical rhetorical frames embedded within a specialized corpus of 800 texts (871,325 tokens). By isolating a persistent lexical core, the proposed method excavates underlying discursive dimensions at a meso-level, successfully bridging the analytical gap between micro-level stylistic features and macro-level thematic topics. The empirical application reveals four stable dimensions: (1) National Strategy & Development vs. Individual & Daily Life, (2) Regional & Domestic Affairs vs. Global & Identity Narratives, (3) Discourse of Retrospective Achievements, and (4) Discourse of Prospective Initiatives. Diachronic tracking demonstrates that the most significant coverage change during this period was primarily a rhetorical adaptation, not a topical one & horbar;specifically, a pivot from a top-down, state-level narrative towards a bottom-up, personal-level storytelling approach. Ultimately, this study contributes a robust, generalizable methodology for diachronic discourse analysis and provides a data-driven, meso-level framework for English for Specific Purposes (ESP) pedagogy, informing instruction that fosters learners' strategic rhetorical competence. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
This paper explores medical discourse as realistically portrayed in audiovisual materials, in order to extract potential pedagogical implications for ESP settings (English for Specific Purposes). In particular, we focus on recontextualization strategies used in medical communication, which can help enhance the communication skills of future healthcare professionals and science communicators. The study takes a communicative perspective and supports using TV series as sources of quasi-authentic interactions that provide examples of realistic, context-rich discourse. Our approach is multimodal, taking into account the wide range of semiotic resources that aid effective communication in medical settings. We perform a Multimodal Interaction Analysis of six selected excerpts of the TV series The Good Doctor, featuring recontextualization strategies. Our results reveal significantly different modal configurations in the two distinct layers of recontextualization found in the materials: recontextualization for characters in the series (e.g. patients and care-givers), which is performed through a combination of verbal and non-verbal resources; and recontextualization for the audience, present in complex unabridged medical interactions, which is mainly achieved through filmic resources. The results can inform pedagogical proposals for EMP courses (English for Medical Purposes) as well as training for Health Care scientists willing to disseminate their findings. (c) 2026 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
In the era of digital humanities, academic writing has undergone a significant shift toward multimodality, with increasing integration of visual elements, such as figures and tables, into scholarly discourse. Although well-documented, this shift has left the linguistic mechanisms of verbal-visual cohesion, particularly in the form of graphical markers, insufficiently explored. This study examines how these markers are realized lexicogrammatically and recontextualized diachronically in applied linguistics research articles across a 45-year span (1981-2025). The analysis focuses on three dimensions: surface form (integral vs. non-integral), grammatical role, and verb choice. The results reveal a robust long-term increase in non-integral markers, with a significant acceleration in the later periods. Within integral forms, adjunct roles decline, while subject realizations become more prominent and object occurrences increase in later periods. Verb choices remain dominated by research-act verbs, but discourse-act verbs rebound in later periods. These findings establish graphical markers as evolving rhetorical resources for guiding readers and navigating evidence in academic writing. They also extend genrebased multimodal research by operationalizing the verbal-visual interface from a diachronic perspective. In addition, the study underscores the pedagogical importance of helping academic writers deploy graphical markers strategically. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Three minute thesis (3MT) presentations challenge doctoral students to communicate scientific research effectively to non-specialists within a strict 3-min limit. Its judging criteria 1 highlight the dual demands of effective science communication: comprehension (fostering the audience's epistemic understanding) and communication (building relational connection with the audience). To meet these requirements, speakers often employ involvement strategies across both dimensions to bridge the gap between science and audience. However, existing 3MT research predominantly examines linguistic realization of these dimensions in isolation, leaving a gap for an integrated framework that explains how strategies are employed together for effectiveness. Drawing upon intersubjectivity, this study develops such an analytical framework and applies it to analyze a corpus of 53 award-winning medical presentations (24,971 words) through a mixed-methods approach using the UAM corpus tool. Findings reveal that in the epistemic dimension, speakers tend to involve audiences by metaphorization to simplify concepts and intensification to strengthen arguments; in the relational dimension, speakers are inclined to involve audiences by non-conducive questions, implicit directives and communitization statements to build rapport and shared atmosphere. The study not only contributes an integrated framework for examining involvement in spoken academic discourse but also offers practical discursive strategies for ESP and science communication pedagogy to enhance audience participation. (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
