---
name: journal-article-development
description: >
  Turns completed research, or a thesis chapter, into a journal article built on a
  single defensible contribution claim. Use for "turn my chapter into a paper",
  "write a journal article", "what is the contribution of this paper", "my paper
  is too long", "how do I structure a research article", "is this publishable
  yet", "who should be an author on this", "write the abstract", "the reviewers
  said there was no clear contribution".
category: 15 Academic University Research
ref: "15.24"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Journal Article Development

## 1. One-line description

A method for converting completed research into a journal article that stands alone on a single, explicitly stated contribution, structured to the conventions of the author's own field, with authorship and publication-readiness settled honestly before submission.

## 2. What this skill is used for

**The research problem it solves.** Most unsuccessful article submissions are not rejected for weak data. They are rejected because the paper does not say what it adds, or says it in the last paragraph, or makes three half-claims instead of one whole one. The commonest source of this failure is a thesis chapter lightly edited into article form. A chapter is written to serve a larger document: it can defer, it can assume what an earlier chapter established, it can survey exhaustively because the examiner is checking coverage, and it can carry material that earns its place by demonstrating competence rather than by advancing an argument. An article has none of those permissions. It is read once, by a reviewer who owes it nothing, alongside a pile of others, and it must be self-contained and single-minded. This skill installs the discipline that transformation requires: naming the one contribution, cutting everything that does not serve it, and rebuilding the architecture around it.

**Where it sits in the research lifecycle.** After analysis and interpretation are complete, and before venue selection and submission. It also runs after a rejection, where the paper needs restructuring rather than a new home.

**Typical use cases.**
- Converting one or more thesis chapters into a standalone article.
- Building an article from a completed study that was never part of a thesis.
- Diagnosing a paper that has been rejected with "no clear contribution" or "reads like a report".
- Cutting an over-long manuscript to a word limit without losing the argument.
- Writing or rewriting a title and abstract that are doing the paper no favours.
- Settling an authorship order and contribution statement before it becomes a dispute.
- Deciding honestly whether a piece of work is publishable yet, or needs more evidence first.

**Who uses it.** Doctoral and postdoctoral researchers publishing from a thesis; early-career academics under publication pressure; established researchers writing outside their usual format or field; research staff in applied settings publishing from commissioned work; supervisors advising on a student's first paper.

## 3. When to use it

- You have a completed piece of research, with analysis finished, and no article yet.
- You have a thesis chapter and an instruction, or an ambition, to publish from it.
- You have a draft that keeps growing and you cannot say in one sentence what it contributes.
- A co-author, supervisor or reviewer has said the paper lacks focus, or has more than one paper in it.
- You are over a word limit and cutting has become arbitrary.
- The paper is written but the abstract and title were produced in a hurry at the end.
- Authorship has not been discussed and people have started to assume different things.
- You need to decide whether to submit now or run one more analysis first.

## 4. When NOT to use it

- **The analysis is not finished, or the findings are unstable.** A paper cannot be built around a contribution that may change when the last model is run or the final interviews are coded. Finish the analysis first: **15.10 Data Analysis Chapter Development** for the results, **15.11 Discussion and Contribution Development** for the interpretation. Writing a paper around a provisional result invites the worst kind of revision, the kind where the argument has to be rebuilt after the data moves.
- **There is no contribution, only a competent piece of work.** A study can be methodologically sound, honestly reported, and add nothing anyone needed to know. That is a real outcome and this skill will surface it. The honest responses are to find the contribution the work actually makes (often smaller and more specific than the one intended), to combine it with other work into something that does contribute, or to publish it in a venue whose purpose is the record rather than the advance. This skill will not manufacture novelty. See **15.13 Doctoral Research Positioning and Originality** where the originality claim itself is the problem.
- **The question is where to send it, not what it says.** Venue fit, review model, access model and predatory-publisher screening belong to **15.25 Journal and Venue Selection**. This skill takes the paper to a submittable state; it does not select the target. The two interact (a target venue's conventions shape the final draft), so run 15.25 in parallel once the contribution is fixed, but do not let venue choice drive the contribution claim.
- **The paper is already under review, or has been reviewed.** Responding to reviewers is **15.26 Peer Review Response**, which also carries the confidentiality rules that apply to a manuscript in review. Do not restructure a paper mid-review outside that process.
- **The work is a thesis and will stay one.** Coherence across a long document, the golden thread from question to conclusion, and chapter-level architecture are **15.18 Thesis Architecture and Chapter Coherence**. The two methods pull in opposite directions: a thesis is built to be complete, an article to be singular. Applying article discipline to a thesis chapter that must remain a thesis chapter will strip out material the examiner requires.
- **The underlying research has an unresolved ethical or governance problem.** Missing or inadequate ethics approval, consent that does not cover publication, data that cannot be de-identified at the granularity the paper needs, or a participant group that could be harmed by the finding. None of these is fixed by careful writing. Resolve them first: **15.09 Ethics Clearance Application**, and **13.05 Research Ethics and Consent Design**. `RESEARCHER DECISION REQUIRED` (K5 §2.4).
- **What is actually wanted is a passage of text to submit as the author's own unaided writing.** This skill structures, interrogates, critiques and teaches. It does not ghost-write a manuscript. See §12.1.
- **The material is better served by a different output.** A single striking observation with no analytic apparatus behind it is not an article; it may be a short communication, a letter, a registered report proposal, or a conference paper. A large body of related work may be a review or a monograph rather than one article. Forcing material into the full-article format because that is the format available is a common and expensive mistake.

## 5. Required inputs

**Required.**
- **The completed research.** Methods as actually executed, results as actually obtained, and the analysis. Not a plan, not a partial run. If results are provisional, stop (see §4).
- **A statement of what the author believes the paper contributes.** Even a rough one. It is the object the first steps work on, and it is almost always revised. If the author genuinely cannot state one, that is the finding, and Step 1 addresses it.
- **The field or subfield.** Article conventions, reporting standards, expected length, section structure, citation density and the meaning of "contribution" all differ by discipline. Without this, the skill can only offer a generic architecture that will fit no field well. If not supplied, ask. Do not assume.

**Optional, and what each one adds.**
- **The source chapters or manuscript draft.** Lets the work proceed by subtraction and restructuring rather than from a blank page, and makes the cut list concrete.
- **A target venue or a shortlist.** Fixes word limits, structural requirements, whether a structured abstract is required, reference style, and the audience's background knowledge. Without it, draft to the field's modal conventions and flag every venue-dependent choice.
- **The applicable reporting guideline for the design.** Many fields have checklists for specific study types (trials, observational studies, systematic reviews, qualitative research, economic evaluations, prediction models, and others). Where one applies, it converts "did I report enough method?" from a judgement into a checkable list.
- **Two or three recently published articles the author considers exemplary in the target area.** The most efficient way to calibrate structure, length, register and citation practice for a field this skill should not assume it knows.
- **The list of everyone who worked on the study and what each did.** Turns the authorship discussion from a negotiation about status into an application of criteria (Step 9).
- **Any prior publication or preprint from the same dataset.** Determines what must be cited, what would be redundant publication, and what has to be disclosed at submission.
- **Funder, institutional or collaboration requirements.** Open access mandates, data availability requirements, embargo terms, and internal review or clearance obligations that constrain where and how the work can appear.

## 6. Questions to ask before starting

1. **What is the one thing a reader knows after this paper that they did not know before?** The single most useful question in the process. If the answer takes more than two sentences, the paper is not yet one paper. Default if unanswered: derive candidate claims in Step 1 and put them back to the author for selection.
2. **What is the field, and what does a contribution look like in it?** Empirical novelty, theoretical development, methodological advance, replication, synthesis, or a negative result all count as contributions in some fields and not others. Default: ask for exemplar papers rather than guessing.
3. **Who is the reader: specialists in the subfield, the broader discipline, or an interdisciplinary audience?** Determines how much is assumed, how much is explained, and how the significance is framed. Default: assume subfield specialists and flag every passage that would fail with a broader readership.
4. **Is there a reporting guideline or field standard the methods must meet?** Default: ask. Do not name or invent one; ask what applies in this field and follow it if supplied.
5. **Has authorship been discussed with everyone involved?** Default: assume it has not, because it usually has not, and raise it early rather than at submission (Step 9).
6. **What constraints are fixed?** Word limit, submission deadline, funder mandate, embargo, or a prior commitment to a venue. Default: draft to field-modal length and mark venue-dependent decisions.
7. **Is any part of this already published, submitted, preprinted or under review anywhere?** Determines overlap risk, citation obligations and disclosure duties. Default: ask explicitly, because authors frequently do not think of a preprint or a conference paper as prior publication.

## 7. Step-by-step methodology

**Step 1. Extract the candidate contributions, then choose exactly one.**
List every distinct claim the work could support, as a full sentence with a subject and a verb. "We studied X" is not a claim. "X operates differently under condition Y than the dominant account predicts" is. Expect four to ten candidates from a thesis chapter, two to five from a standalone study. Score each on three questions: is it true given the evidence, is it new relative to the published literature, and would a competent person in the field care. A claim can fail one and still earn a place in the paper as support rather than as the headline. Then select one. Selecting more than one is the failure this step exists to prevent: a paper with two contributions has two introductions competing for the opening, and reviewers read the competition as confusion. Where two candidates are both strong and genuinely independent, the correct outcome is two papers, and saying so here saves months. *Correct result: one sentence, written down, that the rest of the paper serves. Every subsequent decision is tested against it.* `RESEARCHER DECISION REQUIRED` where the choice between candidate contributions is close (K5 §2.1, §2.3); the judgement depends on knowledge of the field's current conversation that the analysis does not contain.

**Step 2. Establish what the field already knows, precisely enough to locate the gap.**
The contribution claim is a claim about a boundary: this is where existing knowledge stops. Locate it by identifying the three to eight works the paper is genuinely in conversation with and stating, for each, what it established and where it stops. A literature list of eighty items does not locate a boundary; it demonstrates reading. Where the boundary cannot be located, the claim is unsafe, and the honest move is back to a fuller search (**15.02 Academic Literature Search Strategy**, or **15.06 Systematic Literature Review**) rather than forward to drafting. *Correct result: a gap statement of two or three sentences naming the works that define the edge of what is known, with the claim positioned against them.*

**Step 3. Cut everything that does not serve the claim, and record what was cut.**
This is the hardest step and the one most often skipped, because everything in a thesis chapter was written for a reason and the reasons are still remembered. Sort every element (each analysis, table, figure, literature section, methodological justification, paragraph of context) into: **serves the claim** (keep), **supports something that serves the claim** (keep, subordinate, compress), **interesting and unrelated** (cut to a holding file for a future paper), and **present to demonstrate competence** (cut). The last category is the largest in thesis-derived papers: methodological throat-clearing, philosophical positioning that does not change what was done, literature surveyed for coverage. Keep the cut file rather than deleting, because it is the second paper and knowing it exists makes cutting psychologically possible. Heuristic: if a table were removed and no sentence needed rewriting, the table was not serving the argument. *Correct result: a retained set that is smaller than the source, and a cut file. Expect a chapter to lose between a third and two thirds of its content.*

**Step 4. Build the architecture, to the field's conventions rather than a generic template.**
The familiar introduction, methods, results, discussion structure is standard in much empirical work and is not universal. Theory papers, humanities articles, methodological papers, case-based work and many qualitative traditions use different architectures, and imposing the wrong one is immediately visible to a reviewer. Establish the field's convention from the exemplar articles rather than assuming, then map the retained material onto it. Write the section-level plan before drafting prose: for each section, one sentence saying what it must accomplish and a list of the elements it carries. Check that every retained element has a home and that no section is doing two jobs. *Correct result: a one-page architecture, each section's job stated and its content assigned, matching published work in the target area.*

**Step 5. Write the introduction to do four jobs in a small number of paragraphs.**
An introduction establishes: what the problem is and why it matters, what is known, where knowledge stops (the gap from Step 2), and what this paper does about it. In most empirical fields that is a handful of paragraphs, not a chapter. Two failure patterns dominate. The funnel that never lands: pages of ever-narrowing context and no statement of what the paper contributes. And the deferred claim, where the contribution appears only in the discussion, on the assumption that a reader will wait for it. They will not; a reviewer decides what kind of paper this is in the first page. State the contribution explicitly, in the introduction, in words close to the sentence from Step 1. Where the field's convention includes a roadmap paragraph, say what each section does rather than announcing that a literature review follows. *Correct result: a reader who has read only the introduction can state the paper's claim, why it matters, and what evidence is coming.*

**Step 6. Make the literature section argue rather than survey.**
A survey lists what has been done, usually chronologically or by author, and leaves the reader to work out the point. An argument organises prior work around the question the paper answers, groups it by position or finding rather than by author, and states what the pattern shows. The test: read each paragraph's topic sentence in isolation. Topic sentences that name authors mean a survey; topic sentences that name positions, findings or tensions mean an argument. Every work cited must be doing a job in that argument, and a citation present only to show it was read belongs in the cut file. Where the literature is integrated into the introduction rather than given its own section, apply the same test there. *Correct result: a section whose topic sentences, read alone, form a coherent case for the gap the paper fills.*

**Step 7. Report methods to the standard the field requires, not to the standard that fits the space.**
The governing principle is reproducibility: enough that a competent reader in the field could repeat the study, and enough that a reviewer can judge whether the conclusions are licensed by what was done. Report what was actually done, including departures from the plan. Where the field has an applicable reporting guideline for the design, work through it item by item and record where each is addressed; where an item does not apply, say why rather than omitting it silently. Where no guideline applies, the minimum set is still specifiable: design; population and sampling with its frame and any refusal or attrition; materials or instruments with their provenance and validation; procedure with timing and setting; analysis with every decision that could have gone another way; and the ethics approval and consent basis. Qualitative work adds the researcher's position and its bearing on the analysis, the analytic approach as actually applied, and how interpretations were checked. Pre-registration, protocol deviations, and data and code availability go here where the field expects them. *Correct result: a methods section against which a reader could reconstruct the study, with every item a reviewer would ask for either present or explicitly marked unavailable and why.*

**Step 8. Present results without interpretation, and discuss without repeating results.**
The most common structural fault in thesis-derived papers is interpretation leaking into results and results being restated in the discussion, which wastes words twice and blurs the line between what was observed and what is argued (K2 §2.1). Results report what was found: the analyses run and the outputs obtained, including those that did not support the expected pattern, with the completeness the field requires (effect sizes and uncertainty intervals alongside test statistics where quantitative; enough data extract to let a reader judge the interpretation where qualitative). Every table and figure is referred to in the text and carries information the text does not simply repeat. The discussion then does four things and only these: states what the results mean in relation to the claim, situates them against the literature from Step 2 (the studies that disagree deserve the most space), states the limitations, and says what follows. Limitations must be specific and consequential: "the sample was limited" is a ritual, while "the sample came from one institution and the effect may not hold where the intake process differs, which is the assumption the second half of the argument rests on" tells a reader how far to trust the claim. Say which limitation would most change the conclusion if it bit. *Correct result: a results section a hostile reader could not accuse of arguing, and a discussion that adds an argument rather than a summary.*

**Step 9. Settle authorship and contributorship explicitly, before submission.**
Authorship criteria vary in wording between fields and publishers but converge: an author has made a substantial intellectual contribution, has been involved in drafting or critically revising, approves the submitted version, and accepts accountability for it. All parts, not one. Two errors are common and both are misconduct in most codes. Omitting someone who meets the criteria (frequently a student, a technician who made intellectual decisions, or a collaborator who left) is the more serious. Including someone who does not (a head of unit, a funder, a colleague added by reciprocal arrangement) is the more frequent. Funding, providing access or data, and general supervision are contributions and, without more, are acknowledged rather than authored. Order conventions differ sharply by field and the same position carries different meaning in different disciplines, so establish the convention rather than assuming it. Write a contribution statement naming what each person did, using the field's or publisher's taxonomy where one exists. Do it early: a contribution statement written at the start is a description, and one written at the end is a negotiation. *Correct result: a written authorship list, order, and per-person contribution statement, circulated to and agreed by everyone named, with the criteria applied recorded.* `RESEARCHER DECISION REQUIRED` on any contested inclusion, exclusion or order (K5 §2.8); this is a professional accountability judgement and cannot be delegated.

**Step 10. Write the title and abstract last, and give them the time they deserve.**
These are the most-read and least-invested parts of any paper. Most people who encounter the work read only these, and editors and reviewers form their first judgement from them. Write the title to state the finding or the specific object of study rather than the activity, in fields whose convention permits it, and avoid the title that could sit on fifty papers. Then write the abstract from the finished paper, not from the plan, checking every claim against the section it summarises. Where a structured abstract is required, fill each field with what it asks for; where it is not, the same content is still required in prose: the problem, what was done, what was found with the actual result, and what it means. The commonest abstract failure is the empty result sentence ("findings are discussed"), which tells a reader nothing and reads as concealment. Every number in the abstract must appear in the paper (K2 §5). *Correct result: a title a searcher would click and a reader would remember, and an abstract that would let someone decide correctly whether to read the paper.*

**Step 11. Judge whether the paper is publishable yet, and say so plainly.**
Assess four things independently. **Is the contribution real?** Write the one-sentence summary a reader would give a colleague, and ask whether that sentence is worth their attention. **Is it supported?** Trace the claim back through the discussion, the results and the methods, checking that each link holds and that no step carries more confidence than the evidence beneath it (K3 §3.6). **Is it complete?** Every element a reviewer in this field will look for is present, and anything missing is missing because it does not exist rather than because it was inconvenient. **Is it ready?** The paper answers the obvious objections rather than hoping they will not be raised. Where the assessment is negative, name what would fix it and what that costs. An honest "not yet, and here is why" saves a rejection and several months; submitting in the hope that reviewers will do the work has a poor return and a reputational cost. *Correct result: a stated verdict with reasons, and either a submission decision or a specific work list.* `RESEARCHER SIGN-OFF REQUIRED` before submission (K5 §2.8).

## 8. Analytical framework

The paper is built on one structure, applied recursively:

    Contribution claim → Gap it fills → Evidence that establishes it →
    Method that licenses the evidence → Limits of the claim → What follows

Every element in the manuscript occupies a position in this chain, and any element that does not is a candidate for the cut file. The chain is also the order of the reviewer's scrutiny, which is why a weak link anywhere is fatal regardless of the strength elsewhere: a strong result obtained by a method that does not license it establishes nothing, and a well-supported claim that fills no gap is a report rather than an article.

Applying it in practice: take the sentence from Step 1 and walk it left to right through the manuscript. The gap should be visible in the introduction, the evidence in the results, the licence in the methods, the limits in the discussion, and what follows in the conclusion. Anywhere the chain is invisible to a first-time reader, the paper has a structural fault, not a wording fault, and rewriting sentences will not repair it.

A second framework governs the section-level test. For each section, ask: **what job does this do, and would the paper fail without it?** A section that cannot answer the second question is either misplaced material or padding. This test does more work in cutting than any word-count target, because it cuts by function rather than by proportion.

## 9. Output format

**1. Contribution statement.** One sentence. The claim the paper makes, with the candidates considered and rejected recorded beneath it.

**2. Gap statement.** Two or three sentences locating the claim against the works that define the boundary of current knowledge, with those works named.

**3. Architecture plan.** A table.

| Section | Job it does | Content assigned | Approximate length |
|---|---|---|---|

**4. Cut list.** What was removed, and why, split into "belongs to another paper" and "does not earn its place". Retained as a file, not deleted.

**5. The manuscript**, in the field's conventional structure.

**6. Reporting completeness check.** Where a field guideline applies, item by item with the location of each. Where none applies, the minimum method set from Step 7, with any absent item and the reason.

**7. Authorship and contributorship record.** The list, the order, the per-person contribution statement, the criteria applied, and confirmation that everyone named has seen and agreed it.

**8. Title and abstract**, with the alternatives considered.

**9. Publication-readiness assessment.** The four judgements from Step 11, each with a verdict and a reason, and a work list where the verdict is negative.

**When the material is thin, the format must not force a claim (K4 §1).** Where Step 1 produces no candidate that passes all three tests, the correct output is the contribution statement section carrying an explicit negative finding, the candidates with the test each failed, and a specification of what evidence would create a contribution. An honest "this is not yet a paper, and here is what it would take" is a successful output of this skill. Do not build the architecture around a claim the work does not support, and do not pad a thin results section with additional analyses run to fill space, which reviewers recognise and which raises multiplicity problems the paper then has to answer for.

## 10. Quality checks

Run before submission. These sit on top of K4 §8.

1. Can the contribution be stated in one sentence, and does that sentence appear in the introduction as well as the abstract?
2. Does every retained section, table and figure serve the contribution claim, tested by asking what would need rewriting if it were removed?
3. Is the paper self-contained: does it depend anywhere on knowledge the reader would have only from the thesis or a companion paper?
4. Does the literature section argue, tested by reading its topic sentences alone?
5. Could a competent reader in this field repeat the study from the methods section?
6. Where a reporting guideline applies, has every item been addressed or explicitly marked not applicable with a reason?
7. Is the results section free of interpretation, and the discussion free of restated results?
8. Are the results that did not support the expectation reported as fully as those that did (K4 §4.1)?
9. Is every limitation specific, and is the one that would most change the conclusion named as such?
10. Does every number in the abstract appear in the paper, with the same value?
11. Does the abstract state what was found rather than that findings exist?
12. Does every person meeting the authorship criteria appear, and does everyone appearing meet them?
13. Has everyone named seen and agreed the submitted version, the author order and the contribution statement?
14. Is any prior publication, preprint, conference version or overlapping submission disclosed?
15. Does any claim in the paper exceed the confidence its evidence supports, particularly in the abstract and the conclusion (K3 §3.6)?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **The chapter with the edges filed off** | Section headings survive from the thesis; the paper assumes an earlier chapter | Rebuild the architecture from the claim, do not edit the chapter (Steps 3 and 4) |
| **Two papers in one** | The abstract has two "we show that" clauses; the discussion has two halves that do not meet | Choose one contribution and file the other (Step 1) |
| **The deferred claim** | The contribution first appears on page 14 | State it in the introduction (Step 5) |
| **The survey that does not argue** | Literature paragraphs open with author names | Reorganise by position and finding (Step 6) |
| **Interpretation in the results** | The word "because" or "suggests" appears before the discussion | Enforce the K2 finding-to-interpretation boundary (Step 8) |
| **The ritual limitation** | "Sample size was a limitation", with no consequence stated | Name what each limitation would change (Step 8) |
| **The empty abstract** | Ends with "implications are discussed" | Write the abstract from the finished paper, with actual results (Step 10) |
| **Method reported to fit the space** | Analysis decisions absent because the section ran long | Report to the field standard, cut elsewhere (Step 7) |
| **Selective reporting** | Only the analyses that worked appear | Report the full analytic path, including what failed (K4 §4.1) |
| **Authorship settled at submission** | The order is being negotiated the week of the deadline | Apply criteria and write the statement early (Step 9) |
| **Gift and ghost authorship** | A senior name with no involvement; a contributor omitted | Apply all criteria to every person, both directions (Step 9) |
| **Salami slicing** | The same dataset spread across papers with no independent contribution each | Test each paper against Step 1's three questions independently |
| **AI-invented citations** | A perfectly formatted reference for a plausible claim that cannot be found | Nothing enters the reference list unopened (K4 §2.4, §12.2) |
| **AI-smoothed overclaim** | Hedged findings become confident sentences during redrafting | Re-check every claim against its evidence after any rewrite (§12.3) |
| **Submitting to see what happens** | No readiness assessment was made | Run Step 11 and act on the verdict |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated.

1. **Academic integrity is the binding constraint on everything in this skill.** These skills assist a researcher's thinking, structure and rigour. They do not produce work to be submitted as the researcher's own unaided output. The user must comply with the policies of their institution and of the venue they are submitting to, including any requirement to declare AI assistance. Many publishers now require disclosure of AI use in manuscript preparation and prohibit it entirely in peer review. The skill never writes a passage for submission as though the author wrote it; it interrogates, structures, critiques and teaches. In practice here: propose architectures, test claims against evidence, identify what is missing, critique drafts the author has written, explain why a passage fails, and show the author how to fix it. Do not hand over finished prose to be pasted into a manuscript under the author's name.

2. **Never supply a citation from model knowledge, and never complete a partial one.** Every reference in the paper is something the author has opened. Do not offer a plausible reference as a starting point, do not infer a missing year, volume, page range or identifier, and do not cite a real work for a claim it does not make. A fabricated reference in a submitted manuscript is a misconduct finding against the author, not against the system that produced it.

3. **Never strengthen a claim in the course of improving the prose.** Rewriting for clarity systematically removes hedges, because hedged sentences read as weaker writing. In a research paper the hedge is often the accurate part. After any rewriting pass, re-check each claim against the evidence it rests on and restore the qualification the evidence requires (K3 §4.2).

4. **Never generate a result, a number, an effect size, a participant count or a quotation.** If the manuscript needs a figure that was not calculated, write `[not available]` and say what analysis would produce it. This includes plausible-looking placeholder values inserted to complete a table.

5. **Never manufacture a contribution.** Where the work does not support a novel claim, say so. Reframing a familiar finding in unfamiliar vocabulary to make it look new is the specific failure this rule prohibits, and reviewers in the field detect it reliably.

6. **Never describe a method that was not used, or omit a departure from the plan.** This includes tidying a messy analytic history into a clean sequence, presenting an exploratory analysis as though it had been planned, and describing a sample as representative when its recruitment does not support that.

7. **Do not assume a discipline, a country, a career stage or a publication culture.** Article structure, contribution norms, authorship order conventions, expected length, citation density and the role of preprints all vary enormously between fields. Where a convention matters and is unknown, ask, or state the assumption at the point it bites and mark it as unverified.

8. **Never write the ethics or approval statement from inference.** Approval bodies, reference numbers, consent wording and data governance arrangements are matters of record. Report what the author supplies, and mark anything absent as absent.

9. **Where a decision affects authorship, disclosure or research integrity, mark it for the researcher rather than resolving it.** These carry professional and institutional consequences that sit with a named person (K5 §2.8).

## 13. Best-practice principles

- **A chapter and an article differ by argument, not by length.** A 6,000-word chapter cut to 6,000 words of article is still a chapter. The transformation is structural: one claim, self-contained, everything subordinate.
- **The cut file makes cutting possible.** Nobody deletes work they spent a year on. Moving it to a named file for a future paper removes the resistance, and the file is genuinely useful.
- **Write the contribution sentence before anything else, and revisit it after every major edit.** Papers drift. The sentence is the instrument that detects drift.
- **Reviewers read the abstract, the figures and the limitations first.** Invest accordingly. The parts written last with the least energy are the parts read first with the most attention.
- **The studies that disagree with you deserve more space than the ones that agree.** Engaging seriously with contrary work is the strongest signal of a competent paper, and ignoring it is the most reliable way to draw a hostile review from the author of the work you ignored.
- **Specificity in the limitations section buys credibility everywhere else.** A reader who sees an author name a real weakness precisely believes the rest of the paper more, not less.
- **Report the analysis you actually ran, in the order you ran it.** A clean narrative of a messy analysis is a misrepresentation of method, and it is increasingly detectable through data and code availability.
- **Settle authorship in writing at the start of the work.** Almost every serious authorship dispute traces to an assumption nobody stated while the work was going well.
- **A negative or null result is publishable in fields that value it, and dressing it as something else is not.** Establish whether yours is such a field before deciding the finding is unpublishable.
- **The title is a retrieval device and a memory device.** It has to survive being read in a list of search results and being recalled by someone who read the paper once.
- **Read three recent articles in the target area before drafting, with attention to structure rather than content.** Twenty minutes of this replaces a great deal of guessing about a field's conventions.
- **"Not yet" is a legitimate and often correct verdict.** The cost of a premature submission is not just the rejection; it is the several months and, in some venues, the loss of that venue as an option.

## 14. Worked example

Generic fictional scenario, academic (education research).

**INPUT**

A doctoral researcher has a thesis chapter of about 14,000 words reporting a mixed-methods study of how first-generation university students use academic support services. The chapter contains a survey of 480 students, 22 interviews, a literature review covering three separate bodies of work, and a lengthy methodological justification of the mixed-methods design. The supervisor has said "there is a paper in this". The researcher's stated contribution is "the study contributes to understanding of first-generation student experience".

**PROCESS**

*Step 1.* The stated contribution fails immediately: it names a topic, not a claim. Extraction produces six candidates, including: (a) first-generation students use support services at a lower rate; (b) the gap is not explained by awareness, which was equivalent across groups; (c) interview accounts locate the barrier in a perception that support is for students who are failing; (d) the mixed-methods design demonstrates value here; (e) an existing framework of help-seeking does not account for the identity cost participants describe; (f) service redesign should decouple support from remediation.

Tested against true, new and would anyone care: (a) replicates well-established work; (d) is a claim about the researcher, not the world; (f) is an implication. (b) and (c) together are true and interesting. (e) is the strongest formulation, because it states the same material as a claim against a named theoretical position, which is what makes it a contribution rather than a description.

*The judgement call.* (c) alone would make an easier, purely qualitative paper, written faster. (e) is more ambitious and needs the quantitative result to establish that awareness is not the explanation, which is what licenses the identity account. The field values theoretical engagement, and (c) without (b) invites the reviewer question "is this just awareness?". (e) is chosen, with (b) and (c) as its evidence, (a) noted as replication support and (f) reserved for the discussion.

*Step 2.* The gap is located against four works: two establishing the participation gap, one presenting the help-seeking framework the paper contests, one qualitative study in an adjacent context that identified stigma but did not connect it to the framework.

*Step 3.* The three-body literature review collapses to one, organised around help-seeking. The mixed-methods justification drops from 2,100 words to 300, because the field does not require the defence. Eleven of nineteen tables go: seven demographic, four bearing on nothing in the claim. Two interview themes on accommodation and transport move to the cut file as the seed of a second paper. Retained material is roughly 40 per cent of the chapter.

*Steps 5 to 8.* The introduction is rebuilt to reach the claim in four paragraphs. The literature section is reorganised from "Smith found, Jones found" into three positions on why help-seeking gaps persist, with the claim positioned against the third. The methods section grows rather than shrinks, because interview sampling and analytic procedure were under-reported where the examiner knew the study. Results lose the interpretive sentences carried over from the chapter's discussion, and the discussion loses a page of restated results and gains an engagement with the one published study pointing the other way.

*Step 9.* The supervisor contributed to the design and revised critically: meets the criteria. A second academic who suggested the topic and read one draft, and a research assistant who transcribed and made no analytic decisions: acknowledged. A departmental head who expects inclusion on all unit outputs: does not meet the criteria, and this is raised as a decision for the researcher and supervisor rather than resolved by the skill.

*Step 11.* The contribution is real and supported. One gap surfaces: the identity-cost claim rests on interview accounts alone, and the survey contains an unanalysed item on perceived stigma that could corroborate it. The assessment is "not yet, one analysis short", and that analysis takes two days.

**OUTPUT**

A manuscript of about 7,000 words with one contribution claim stated in the introduction and tested in the discussion, a cut file containing the second paper, an agreed authorship list with a contribution statement, an abstract reporting the actual result, and a readiness verdict that delayed submission by two days and removed the most likely reviewer objection.

`RESEARCHER SIGN-OFF REQUIRED` on the authorship list and on submission (K5 §2.8).

## 15. Advanced usage

**Planning a paper before the study runs.** The strongest use of this skill is prospective. Writing the contribution sentence and the gap statement before data collection surfaces, early and cheaply, the studies whose design cannot support any claim worth making. It also produces the introduction, most of the methods and the analysis plan in advance, which reduces analytic drift and makes pre-registration straightforward where the field uses it.

**Planning a publication set from a thesis.** Where a thesis holds material for several papers, plan them together rather than sequentially. Assign each chapter's material to a paper, check that each has an independent contribution (not a slice of one, which is salami publication), and sequence them so the paper establishing the shared method or dataset appears first. Works with **15.18 Thesis Architecture and Chapter Coherence** in reverse: that skill assembles, this one disassembles.

**Restructuring after a rejection on contribution grounds.** Where reviewers report no clear contribution, the fault is almost never in the data. Re-run Steps 1 to 4 from scratch, ignoring the existing structure entirely, and expect the selected claim to change. Rewriting the introduction rarely fixes this, because the structure beneath it was built for a different claim.

**Multi-author and large-collaboration papers.** Where authorship runs to dozens or hundreds, contribution statements, order conventions and approval mechanisms are governed by the collaboration's own agreement rather than by individual negotiation. Establish that agreement and apply it; do not import the small-team model. Internal review before submission may also be a formal stage with its own timeline, and the readiness assessment in Step 11 belongs before it.

**Where the standard approach does not fit.** Humanities articles in many traditions have no separable methods or results sections, and the contribution may be an interpretation rather than a finding; the claim discipline in Steps 1 to 3 still applies, and Steps 7 and 8 are replaced by the field's own standards of evidence and citation. Registered reports invert the sequence: the contribution claim and method are reviewed before data collection, which makes Steps 1, 2 and 7 the whole of the first submission.

## 16. Skill chain

**Recommended previous skills:**
- **15.11 Discussion and Contribution Development.** Hands over the interpretation and the first articulation of what the work adds, which Step 1 stress-tests and narrows to one claim.
- **15.18 Thesis Architecture and Chapter Coherence.** Hands over the completed thesis material, and the record of which chapters carry which arguments.
- **15.10 Data Analysis Chapter Development.** Hands over results reported to academic standard, which Step 8 reorganises rather than re-analyses.
- **15.13 Doctoral Research Positioning and Originality.** Hands over a defended originality claim where the paper's contribution is contested territory.

**Recommended next skills:**
- **15.25 Journal and Venue Selection.** Takes the finished manuscript and its contribution claim and matches them to a venue, and screens the shortlist.
- **15.26 Peer Review Response.** Takes over once the paper is submitted and reviews arrive.
- **15.27 Conference Abstract and Presentation.** Takes the same contribution claim into a conference submission, often before the article is complete.

**Runs well alongside:**
- **15.03 Referencing and Citation Management**, which verifies every reference the manuscript carries.
- **15.05 Academic Writing Structure and Argumentation**, for paragraph-level argument inside the architecture this skill sets.
- **13.02 Source and Citation Verification**, which audits the reference set before submission.
- **13.06 AI Research Governance**, which covers the disclosure obligations that attach to AI-assisted manuscript preparation.

---
A Yazi Supplied Skill and resource.