Scholarly Writing for Researchers

Scholarly writing is the craft of turning research into a manuscript, proposal, or dissertation that a journal, funder, or examining committee will accept — distinct from the publishing and dissemination systems covered on CASRAI’s Scholarly Publishing hub. That neighbouring cluster covers what happens to a manuscript once it enters the editorial and distribution pipeline: peer review as a system, open access models, preprint servers, research assessment. This cluster covers what a researcher actually has to write, and to the conventions a research office is expected to know when it advises on a manuscript, a grant narrative, or a dissertation: structure, submission mechanics, reviewer responses, style, and the integrity rules that govern reused or AI-assisted text. It is not generic writing help — CASRAI is not in the business of teaching grammar or sentence construction — and it is not career or grammar coaching aimed at a general audience. Every topic here is scoped to a document type a working researcher or research administrator actually produces, and every claim about a rule or convention is checked against the standards body that actually sets it: ICMJE for manuscript structure and authorship, COPE for text-reuse and integrity questions, and NIH/funder guidance for proposal narratives.

What this cluster covers

Four broad categories of document sit inside this cluster’s scope:

• Journal manuscripts — the structure, front matter, and submission mechanics of an original-research article, from first draft to a decision letter.
• Reviewer correspondence — decision letters, response-to-reviewers documents, and the revision process that follows peer review, from the author’s side.
• Grant and fellowship narratives — specific aims pages, research strategy sections, and the narrative CV formats (biosketch, R4RI) that increasingly replace list-format CVs in funding applications.
• Theses and dissertations — the long-form research document itself; see CASRAI’s dissertation structure and deposit guide for the doctoral-degree mechanics this cluster doesn’t repeat here.

Running through all four is a fifth concern that doesn’t have its own document type but touches every one of them: writing integrity — authorship, text recycling, salami publication, and (new since 2023) generative-AI disclosure. A research office that advises on any of the four document types above needs a working grasp of where the integrity line sits, so it’s treated as a full section below rather than a footnote.

The anatomy of a manuscript: IMRAD and its variants

Most original-research articles in the sciences, and a large share in the social sciences, follow the IMRAD structure — Introduction, Methods, Results, and Discussion — a convention the International Committee of Medical Journal Editors (ICMJE) describes as reflecting “the process of scientific discovery” rather than an arbitrary house style. ICMJE’s own recommendations, followed in some form by thousands of journals beyond medicine, specify each section’s job:

• Introduction — establishes context, states the specific purpose or hypothesis, and cites only directly relevant prior work rather than a full narrative review of the field.
• Methods — ICMJE’s guiding principle is “clarity about how and why a study was done in a particular way,” detailed enough that another researcher could assess or repeat it — the same reproducibility bar CASRAI’s Research Data Management cluster covers from the data side.
• Results — presents findings in logical sequence without duplicating what a table or figure already shows.
• Discussion — summarizes the main findings, states limitations explicitly, and distinguishes statistical from practical significance.

Not every document uses IMRAD. A protocol paper describes a planned or ongoing study rather than reporting completed results, and is structured around methods and planned analysis instead. A Registered Report splits the same content across two review stages — a Stage 1 protocol (introduction and methods, reviewed and provisionally accepted before data collection) and a Stage 2 completed manuscript — specifically to prevent the results from influencing whether the study gets published at all. A narrative review, by contrast, has no methods/results split at all; it’s organized thematically. Knowing which structure a given submission calls for is often the first thing that goes wrong in a first draft.

Abstracts, titles, and the front matter that gets a paper read

ICMJE requires a structured abstract for original research, systematic reviews, and meta-analyses: a stand-alone summary giving “the context or background for the study,” its purpose, basic procedures, main findings — “giving specific effect sizes and their statistical and clinical significance, if possible” — and principal conclusions, plus (where applicable) a clinical trial registration number and any dataset identifiers. The title page carries its own conventions: full author list with institutional affiliations, a corresponding author’s contact details, a word count that excludes the abstract/references/tables/figures, and any funding disclaimers — some journals cap the title itself at around 40 characters. A conference abstract is a distinct, shorter genre with its own submission deadlines and formatting rules, not a draft of the eventual journal abstract.

Two further front-matter items sit outside the manuscript body but are part of the same submission package: the cover letter to the editor (a short case for fit and significance, and, per ICMJE’s authorship guidance, the place to disclose any generative-AI assistance used in preparing the manuscript) and a data availability statement describing where the underlying data can be found — now required by most major publishers and funders at submission, not after acceptance.

A structured abstract’s four to five labelled parts (background, methods, results, conclusions — sometimes with a separate objective/purpose line) exist so an editor, reviewer, or database indexer can assess a paper’s contribution without reading the full text — which is also why ICMJE asks specifically for effect sizes and their statistical or clinical significance rather than a vaguer “results were significant” summary. Keywords, usually five to eight, are the other piece of front matter that does real indexing work: they determine how a paper surfaces in database searches independent of its title, so generic or overly broad keywords (repeating title words rather than adding new retrieval terms) are a common, easily fixed submission weakness.

Journal submission mechanics: choosing a venue, author guidelines, and cover letters

Venue selection has real, verifiable stakes beyond fit: a growing share of “journals” soliciting submissions are predatory journals that charge a fee without providing genuine peer review or archiving — CASRAI’s guide on how to identify predatory journals and publishers covers the checklist a research office should run before a researcher submits. Once a legitimate venue is chosen, “author guidelines” (sometimes “instructions for authors”) set binding, journal-specific rules — word limits, reference style, figure formats, and required disclosure forms — that sit on top of, not instead of, ICMJE’s general recommendations. Nearly every modern journal now routes submission through a manuscript-tracking system rather than email; CASRAI’s Editorial Manager vs. ScholarOne vs. OJS comparison covers how the three most widely used systems differ for an author or editorial office.

Two disclosures belong in the submission package itself, not as an afterthought: a conflict of interest disclosure (see CASRAI’s guide on building a conflict of interest disclosure form) and, where relevant, confirmation of IRB or equivalent ethics review — CASRAI’s IRB/REC approval process guide covers what does and doesn’t need it. Many researchers also deposit a preprint before or alongside formal submission; see the Scholarly Publishing cluster’s guide on choosing a preprint server for that adjacent decision.

Responding to peer review: revisions, rebuttals, and resubmission

ICMJE’s own framing of peer review is worth stating plainly to authors who treat a “reject” or “major revision” decision as a verdict on the science: ICMJE describes peer review as valuable but “not perfect” and explicitly not “beyond dispute.” Editors are responsible for selecting reviewers competent to evaluate a given manuscript, and for giving authors a genuine opportunity to respond to reviewer criticism before a final decision — see CASRAI’s guide on the peer reviewer role and the general peer review guide for how the process looks from the other side of the desk.

A response-to-reviewers document — sometimes called a rebuttal letter — is the actual deliverable a research office can help draft or check: it should address every point raised, point by point, state plainly whether each change was made or why it wasn’t, and reference the specific location of each change in the revised manuscript. A decision letter after review typically falls into one of a few standard categories — accept, minor revision, major revision, reject-and-resubmit, or reject — and each calls for a different response: a minor-revision letter can usually be answered directly point by point, while a major-revision letter often benefits from a short cover note up front summarizing the overall approach to the reviewers’ concerns before the detailed point-by-point response. Where a reviewer’s request would require new experiments or data collection that genuinely can’t be done in the resubmission window, ICMJE’s own framing of peer review as imperfect and disputable is the legitimate basis for a reasoned, respectful pushback in the response letter — not a reason to concede a point that isn’t actually correct. Post-publication, the same discipline applies to a different document: CASRAI’s post-publication peer review guide covers what happens when a concern is raised after a paper is already out, including expressions of concern and corrections.

Academic style and clarity conventions

Style conventions vary meaningfully by discipline and journal — there is no single “academic style” — but a small set of cross-disciplinary conventions recur across STEM and social-science venues: consistent tense (methods and results typically past tense, established facts present tense), first mention and definition of every abbreviation, metric units, and precise hedging language that distinguishes what the data show from what the authors infer. Citation style is the one convention every manuscript has to get exactly right, because it’s mechanically checked at submission: CASRAI’s citation styles and referencing formats guide covers the major formats (APA, MLA, Chicago, Vancouver/NLM, IEEE) and when each applies. Reference accuracy matters beyond formatting, too — ICMJE’s recommendations specifically call for verifying every citation against the original source rather than copying a reference list forward from a prior draft, a real source of the citation errors that later require a correction.

Hedging language is worth calling out specifically because it’s where most non-native and early-career authors lose precision without realizing it: “results suggest,” “may indicate,” and “is associated with” are not interchangeable with “results prove,” “demonstrates,” or “causes,” and a reviewer trained on IMRAD conventions will read the stronger verb as a claim the study design may not support — particularly for observational or correlational work. The same discipline that keeps a Discussion section honest about limitations (see above) should govern verb choice throughout the Results and Discussion, not just the explicit limitations paragraph. This is also where two integrity-adjacent practices covered in the next section actually originate in the drafting process: overstating a finding’s certainty is a milder version of the same instinct that produces HARKing when taken further.

Writing integrity: authorship, plagiarism, text recycling, and AI disclosure

This is the section every other one in this cluster eventually points back to, because the same handful of integrity rules apply whether the document is a manuscript, a grant narrative, or a dissertation chapter.

Authorship is governed, for most biomedical and many other journals, by ICMJE’s four-part test: substantial contribution to the work’s conception/design or data acquisition/analysis/interpretation; drafting the work or revising it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all parts of the work. ICMJE places the responsibility for applying that test on the author group itself, not the journal. CASRAI’s CRediT & Authorship hub covers the companion contributor-role taxonomy in depth; the CRediT vs. ICMJE comparison covers exactly where the two frameworks diverge (CRediT documents specific contributions, ICMJE gates who qualifies for the byline). Disputes over author order are common enough to have their own guide: resolving authorship order disputes.

Plagiarism — presenting someone else’s words or ideas as one’s own without attribution — is the clearest-cut integrity violation; CASRAI’s plagiarism term covers the baseline definition. A narrower, more contested case is self-plagiarism (also called text recycling): reusing one’s own previously published wording. The Committee on Publication Ethics (COPE) frames it as “reusing one’s own previous writing without being transparent about this or appropriately referencing/quoting from the original” — the operative test is disclosure, not reuse itself. COPE explicitly treats minor, boilerplate reuse (a methods description, for instance) as more tolerable than recycled results or discussion text, and recommends a correction for minor breaches versus retraction where the overlap is substantial and undisclosed.

A related practice is salami slicing — splitting one dataset or study into the “smallest publishable unit” across multiple papers to inflate publication count — which COPE defines alongside duplicate publication: publishing “whole or substantial parts of work/data/analysis that have already been published… without transparency or appropriate declaration/referencing.” Both practices distort the literature (they can skew a later systematic review that double-counts the same underlying data) and both turn on the same fix: authors disclosing and citing the overlap rather than presenting each slice as fully independent. A separate, well-documented distortion of the scientific narrative itself is HARKing — presenting a post hoc finding as though it had been hypothesized in advance — which the Introduction and Discussion sections above are specifically structured to prevent.

The newest addition to this list is generative-AI disclosure. ICMJE’s current position, updated in its authorship recommendations, is unambiguous on one point: “chatbots (such as ChatGPT) should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work,” and cannot agree to be accountable for it — the same fourth criterion that gates human authorship. Where AI tools are used for writing assistance, ICMJE expects disclosure — typically in the acknowledgments or methods section and in the cover letter — describing how the tool was used, with the human authors remaining fully responsible for verifying the output, since, in ICMJE’s own words, “AI can generate authoritative-sounding output that can be incorrect, incomplete, or biased.” CASRAI’s generative-AI disclosure statement and AI tool disclosure terms cover the mechanics; the AI co-authorship rejection (ICMJE 2023) term covers the policy’s origin, and acknowledgement vs. authorship for AI covers where the line actually falls. CASRAI’s journal and publisher policy landscape on generative AI in manuscripts guide tracks how individual publishers have implemented this differently, and AI-assisted writing as a term covers the drafting use case specifically, distinct from AI-assisted data analysis.

Grant narratives, biosketches, and dissertations

Outside the manuscript, the same writing craft applies to two other document families a research office routinely reviews. Grant and fellowship narratives have their own, often stricter, formatting discipline: NIH’s Specific Aims attachment, for example, is hard-capped at one page regardless of activity code, ahead of a Research Strategy section capped at 12 pages for a standard R01 — limits set directly in NIH’s own application guide, not a norm authors can negotiate around. Within that single page, NIH’s own guidance points reviewers toward a standard shape: a short framing of the problem and long-term goal, the project’s overall objective and central hypothesis, then typically two to four discrete, independently assessable aims, each stated in a sentence or two along with the approach and expected outcome — written so that a reviewer who has read only this page, not the full Research Strategy, already understands what is being proposed and why it matters. The CV format that accompanies a proposal has also shifted: alongside the traditional list-format biosketch (see the NIH-style biosketch specifically), a growing number of funders — UKRI among them — now require a narrative CV in the R4RI (Résumé for Research and Innovation) format, which asks for a contribution-based narrative rather than a chronological list. CASRAI’s NIH biosketch vs. UKRI R4RI comparison covers exactly how the two formats diverge, and the grants-management cluster’s budget justification narrative guide covers the companion financial-narrative document that goes alongside the science.

Theses and dissertations are the longest-form document in this cluster and have their own defense and deposit mechanics that this pillar doesn’t duplicate — see CASRAI’s dedicated dissertation structure, defense, and deposit guide for the full treatment, including what happens administratively after deposit (embargoes, DOI assignment, and publishing derivative articles from dissertation chapters).

Where this cluster connects: publishing, research integrity, and CRediT

Scholarly writing rarely stays inside its own lane. A manuscript’s authorship list feeds directly into CASRAI’s CRediT & Authorship cluster, where each author’s specific contribution gets documented against the fourteen CRediT roles rather than just the byline order — see CRediT author statement samples for how that documentation actually reads on a published paper, and how to disclose AI assistance in a CRediT statement, role by role for where AI-assisted drafting fits into that same taxonomy. Once a manuscript is accepted, it hands off entirely to CASRAI’s Scholarly Publishing hub — peer review as an editorial system, open access licensing, embargoes, and how the published article is indexed and assessed. And every integrity question raised in the writing-integrity section above — plagiarism, undisclosed reuse, disputed authorship — is adjudicated through the mechanisms covered on CASRAI’s Research Integrity & Compliance hub, including COPE’s own case-resolution flowcharts (see COPE flowcharts explained) and the broader COPE guidelines those flowcharts sit inside. Persistent identifiers thread through nearly every document type here too — a manuscript’s DOI and an author’s ORCID iD connect this cluster to CASRAI’s Persistent Identifiers & Research Information Systems hub.

Frequently asked questions

Is “scholarly writing” the same as academic writing help or grammar coaching?

No. This cluster is scoped to the structural, procedural, and integrity conventions of specific research documents — manuscript structure, submission mechanics, reviewer responses, grant narratives, dissertations — not sentence-level grammar, style coaching, or general essay writing. A candidate topic without a clear researcher/manuscript/journal/reviewer angle sits outside this cluster’s scope by design.

What’s the difference between this cluster and CASRAI’s Scholarly Publishing hub?

This cluster covers what a researcher writes and how; the Scholarly Publishing cluster covers what happens to it afterward as an editorial and dissemination system — peer review as a process run by journals, open access models and embargoes, preprint infrastructure, and research assessment. The two overlap at the manuscript’s edges (submission mechanics, peer review responses) rather than duplicating each other.

Do I need to disclose AI writing assistance, and how?

Per ICMJE’s current recommendations, yes: an AI tool cannot be listed as an author, and its use in preparing the manuscript should be disclosed — typically in the acknowledgments or methods section and referenced in the cover letter — describing specifically how it was used. The authors remain fully responsible for verifying the accuracy and originality of anything the tool produced. See CASRAI’s generative-AI disclosure statement term and the publisher policy landscape guide for how individual journals have implemented this.

Is text recycling (“self-plagiarism”) actually against the rules?

It depends on degree and disclosure, not on reuse alone. COPE treats minor, boilerplate reuse (a standard methods description, for example) as broadly tolerated, while undisclosed reuse of substantial results or discussion text across papers is treated as a genuine integrity breach — up to retraction where the overlap is extensive and undisclosed. See CASRAI’s self-plagiarism and salami slicing terms for the closely related practice of splitting one study into multiple minimal papers.