ESL writers and the Swales / Hyland tradition

Last updated: 2026-05-17

Synthesis

A substantial fraction of scholarly writing today is produced by authors for whom English is an additional language, and the dominant pedagogical tradition for teaching them — running from John Swales’ genre analysis through Ken Hyland’s corpus-linguistic work on metadiscourse — provides the most empirically grounded account of what makes academic prose function as academic prose. The point is not that ESL writers need a softened “easy English” version of scientific writing. The point is that the patterns native speakers acquire by osmosis (hedging force, citation integration, stance marking, interactive vs. interactional metadiscourse) are explicit, conventional, and discipline-specific — exactly the kinds of patterns an agentic editor can audit. The commercial tools that target this audience (Paperpal, Trinka, Writefull) are essentially Hyland’s framework operationalised.

For scriptorium, the load-bearing implication is that ESL-relevant audits — especially hedging calibration, stance/voice consistency, and citation-integration patterns — are not a separate “ESL pass” but the core of what argumentative-flow should already be doing. The Swales/Hyland literature gives us a vocabulary for naming what good sentence-level critique attends to. Framing these checks as “ESL-friendly” rather than “for ESL writers” is the right move: they help every author, but they help non-native writers most.

Evidence and frameworks

Swales’ CARS model

John Swales’ 1990 Genre Analysis: English in Academic and Research Settings [1] introduced the Create-A-Research-Space (CARS) model of research article introductions. The three rhetorical moves are:

1. Establishing a territory — claiming centrality, making topic generalisations, reviewing previous research.
2. Establishing a niche — counter-claiming, indicating a gap, question-raising, or continuing a tradition.
3. Occupying the niche — outlining purposes, announcing present research, announcing principal findings, indicating structure.

CARS is the single most-cited framework in academic-writing pedagogy and is the implicit skeleton of most “how to write an introduction” advice. Swales’ 2004 Research Genres [2] extends the framework beyond introductions and explicitly addresses cross-disciplinary variation; see also discipline-conventions.

The framework matters here because the gap statement is the move that most often goes wrong: too aggressive (“no prior work has…”), too vague (“more research is needed”), or absent. CARS gives an editor concrete questions to ask of an introduction — which move is present, which is missing, which is misordered.

Hyland’s metadiscourse and hedging

Ken Hyland’s work is the empirical complement to Swales’ rhetorical analysis. His 1998 Hedging in Scientific Research Articles [3] is the canonical corpus study of how scientists modulate the strength of their claims through modal verbs (“may”, “might”), epistemic adverbs (“possibly”, “apparently”), approximators (“approximately”, “about”), and indirect attributions (“it has been suggested”). The central finding is that hedging is not optional softening; it is how scientific claims are calibrated to evidence. Under-hedging reads as overclaiming and triggers reviewer hostility; over-hedging reads as evasive or under-confident.

Hyland’s 2005 Metadiscourse [4] distinguishes:

• Interactive metadiscourse — devices that organise the text (transitions, frame markers, endophoric markers, evidentials, code glosses).
• Interactional metadiscourse — devices that signal the author’s stance toward the proposition and the reader (hedges, boosters, attitude markers, self-mentions, engagement markers).

Hyland & Jiang’s 2018 corpus study [5] tracked these features across four disciplines over fifty years and documented a significant increase in interactive features and decrease in interactional types, which they read as academic prose becoming more author-driven and less reader-engaging — a finding that informs how to interpret “stance” advice. The full title — “In this paper we suggest: Changing patterns of disciplinary metadiscourse” — is itself an example of the construction it studies.

Flowerdew on EAL academic writing

John Flowerdew’s work [6] documents the specific obstacles English as an Additional Language (EAL) writers face — not vocabulary or grammar in the surface sense, but the cultural and disciplinary conventions that govern stance, citation integration, and self-positioning. His repeated empirical observation is that the hardest part of EAL academic writing is not producing grammatical sentences; it is producing sentences with the right rhetorical force for the discipline.

Specific patterns ESL writers struggle with

The pedagogical literature [3,4,6] converges on a small set of high-leverage difficulties:

• Hedging miscalibration. Direct translation from a first language can produce claims that are too strong (“this proves”, “this shows clearly”) or too tentative (“it might possibly be suggested that…”). Both fail in characteristic ways under peer review.
• Citation integration. Three patterns coexist in academic English — integral (“Smith (2020) argues…”), non-integral (“(Smith 2020)”), and reporting verbs with varying force (“notes” vs. “claims” vs. “demonstrates”). The choice carries stance information that is invisible to grammar checkers.
• Voice and stance. First-person vs. impersonal constructions, use of boosters (“clearly”, “obviously”), and engagement markers (“note that”, “consider…”) vary by discipline and venue.
• Idiomatic transitions. “Moreover”, “in particular”, “specifically”, “in contrast”, “however” carry slightly different logical loads that are conventionalised, not derivable from dictionary meaning.

These are the things tools like Paperpal and Writefull check; they are also exactly what reader-expectation-approach and argument-mapping should care about at the sentence level.

How this informs scriptorium

1. argumentative-flow should attend to hedging force. A claim in core_claims that the prose hedges (or fails to hedge) inconsistently is a real defect, not stylistic noise. The conservative-edit posture is well-suited: flagging hedging miscalibration is cheaper than rewriting it.
2. Stance consistency belongs in critique skills, not in a separate ESL pass. Branding a check as “ESL” risks ghetto-ising it. Frame it as: “Is the hedging force consistent with the strength of the evidence cited?”
3. Citation integration tells you about stance. A skill that audits citation-claim alignment (citation-claim-alignment) should also flag reporting-verb force — “demonstrates” vs. “suggests” is a hedging choice that should match the cited study’s strength of evidence.
4. Discipline awareness. Hyland & Jiang [5] document disciplinary variation; scriptorium’s knowledge base is biomedical-coded (see discipline-conventions). State the scope.

Implementation priority for scriptorium

Verdict: Maybe later — no dedicated esl-friendliness-pass skill; instead, embed ESL-aware checks in existing critique skills (v0.1 / v0.3) and call this out explicitly in skill descriptions.

If Yes (later): the trigger would be evidence that users want a standalone audit framed around hedging/stance/citation-integration. The skill would be stance-and-hedging-audit, v0.3+. Required data: core_claims and terminology from MANUSCRIPT_STATE.yaml, plus the bibliography for evidence-strength lookup. Output: a list of claims whose hedging force does not match their evidentiary support, with suggested rewordings preserving the citation set.

Why “maybe” rather than “yes now”: the work overlaps heavily with argumentative-flow (v0.1) and any future citation-claim-alignment skill. Spinning it out before those skills mature risks fragmenting responsibility. The right v0.1 move is to make sure argumentative-flow’s description names hedging calibration explicitly so users invoke it for that purpose. Promote to a standalone skill if and only if the embedded version under-uses the framework or users explicitly ask for an ESL-targeted tool.

Condition that flips to Yes: (a) a user requests a standalone hedging/stance pass, or (b) argumentative-flow is shown not to cover hedging force reliably on real manuscripts.

Open questions / weak evidence

• The corpus-linguistic findings are descriptive, not interventional. Hyland tells us what skilled academic prose looks like; whether rewriting toward those norms improves outcomes (acceptance, citation, reviewer sentiment) is largely unmeasured.
• The “ESL writer” category is heterogeneous: a Chinese-L1 computational biologist and a Brazilian-L1 humanities scholar share little. Discipline often dominates language background.
• Commercial tools (Paperpal, Writefull, Trinka) report internal evaluations but rarely publish externally verifiable benchmarks. Their value proposition is mostly an existence proof that the market exists, not evidence that their interventions work.

References

[1] Swales, J. M. (1990). Genre Analysis: English in Academic and Research Settings. Cambridge University Press. ISBN 9780521338134.

[2] Swales, J. M. (2004). Research Genres: Explorations and Applications. Cambridge University Press. ISBN 9780521533348 (paperback; hardback ISBN 9780521825603). [TODO verify hardback ISBN — the originally listed 9780521531344 appears not to be the canonical hardback identifier.]

[3] Hyland, K. (1998). Hedging in Scientific Research Articles. John Benjamins Publishing. ISBN 9789027250698.

[4] Hyland, K. (2005). Metadiscourse: Exploring Interaction in Writing. Continuum. ISBN 9780826476104.

[5] Hyland, K., & Jiang, F. K. (2018). “In this paper we suggest”: Changing patterns of disciplinary metadiscourse. English for Specific Purposes, 51, 18–30. DOI: 10.1016/j.esp.2018.02.001. (The DOI originally referenced as 10.1016/j.esp.2018.04.005 does not resolve to this paper; the correct DOI ends in .02.001.)

[6] Flowerdew, J. (2001). Attitudes of journal editors to nonnative speaker contributions. TESOL Quarterly, 35(1), 121–150. DOI: 10.2307/3587862. See also Flowerdew, J. (2013). Discourse in English Language Education. Routledge. ISBN 9780415499651.