How to write a journal article has two answers: the shape most research articles share, IMRaD, and the rules of the one journal you send yours to. This guide is published by EdCitation, whose tools handle a manuscript's sources, each taken from the publisher's record, and none of its prose. Every rule was read at its source on 24 September 2026. No guide can promise acceptance, and this one does not; it deals with the avoidable faults.
EdCitation's part is the sources: Find sources for the literature, Cite a source for each reference, and Verify references for the finished list, all free with no account. The reference checks style by style are in the journal manuscript references checklist and are not repeated here.
What is IMRaD, and what does each part do?
IMRaD is the order most original research articles follow: Introduction, Methods, Results and Discussion. The International Committee of Medical Journal Editors (2026d) calls it "not an arbitrary publication format but a reflection of the process of scientific discovery": the question, how it was answered, what was found, and what it means.
| Part | Its job, as the ICMJE sets it | The common slip |
|---|---|---|
| Introduction | The problem, why it matters, and the purpose or hypothesis | Data or conclusions from the study itself |
| Methods | How and why the study was done, enough for others with the data to reproduce it; ethics approval; any AI used to do the study | Facts learned during the study, which belong in Results |
| Results | Main findings first, every primary and secondary outcome, absolute numbers beside percentages | Every table repeated in the text |
| Discussion | The findings against all the evidence, the limitations, the implications | Conclusions the data do not support |
| References | Every source, verified against PubMed or the original | An entry no one checked; EdCitation's Verify references checks the list |
Where IMRaD bends
The ICMJE expects subheadings inside the sections, and allows meta-analyses a different format and case reports and narrative reviews a looser one. PLOS ONE (n.d.) fixes the beginning (title page, abstract, introduction) and the end, but lets the middle sections be renamed, reordered, or merged into Results and Discussion. Nature (n.d.) opens an Article with a referenced summary paragraph and puts the Methods after the figure legends. The form is still argued over: Habibzadeh (2026) proposes putting the conclusion first, since readers scanning many papers otherwise meet the main message late.
Where did the IMRaD structure come from?
From medicine's slow adoption of a form recommended long before authors used it. The documented history is Sollaci and Pereira (2004), who took every tenth issue of the BMJ, JAMA, The Lancet and the New England Journal of Medicine at five-year intervals from 1935 to 1985, 1,297 original articles in all.
What the fifty-year survey found
- In 1935 no article was in IMRaD form. The first appeared in the 1940s, and by 1950 they passed 10% in all four journals.
- From 1955 to 1975 the share more than quadrupled, passing 80% in the 1970s.
- The New England Journal of Medicine had fully adopted it by 1975, the BMJ by 1980, JAMA and The Lancet by 1985.
- Early IMRaD papers did not keep today's order: results could come before methods.
The authors trace the paper from the letters of the first journals in 1665, through a "theory, experiment, discussion" pattern in the nineteenth century, to early twentieth-century manuals that already called IMRaD ideal. They found no definite reason for its victory, pointing to physics, which had adopted it widely in the 1950s; to postwar conferences on publishing; to editors insisting on a clear form for readers and reviewers; and to readers who browse for one section. The ICMJE's Uniform Requirements, first published in 1978 to standardise manuscript format, came at the end of that shift (International Committee of Medical Journal Editors, 2026c). The authors warn against extending their four journals' pattern to others.
What could not be confirmed
Writing guides, Robert A. Day's manual among them, credit Louis Pasteur, whose experiments were described closely enough for doubters to repeat them, and trace the methods section to that. No primary source was read, so the story is unconfirmed here, as is the date of 1972 given for an American standard prescribing IMRaD. A library catalogue does list ANSI Z39.16-1979, Preparation of scientific papers for written or oral presentation (Penn State University Libraries, n.d.), but its text could not be read, so what it prescribed is not known here. Day's manual is in EdCitation's Library: its title returned records from 1979 to the ninth edition of 2022, by Barbara Gastel and Robert A. Day.
What does a journal ask of a manuscript?
What its instructions for authors say. Many medical journals also state that they follow the ICMJE Recommendations, last updated in January 2026. Read both before drafting.
What the ICMJE asks for
- A title page with the authors, their institutions, the sources of support, and a word count of the text alone, without the abstract, tables or references (International Committee of Medical Journal Editors, 2026d).
- Authors who meet all four criteria: a substantial contribution, drafting or critical revision, approval of the final version, and accountability for the whole. Chatbots are not listed as authors, and AI used for writing help goes in the acknowledgements (International Committee of Medical Journal Editors, 2026b).
- Trial registration in a public registry at or before the first participant's consent to enrol (International Committee of Medical Journal Editors, 2026a).
- A cover letter or submission form declaring overlapping publication and conflicts, with a statement on authorship; any checklist the journal asks for goes with the manuscript (International Committee of Medical Journal Editors, 2026e).
What journals add
PLOS ONE sets no word limit, and wants double spacing, continuous line numbers, no footnotes and a one-page cover letter. Nature's typical six-page Article runs to about 2,500 words, and its guide says its editors often suggest revised titles and rewrite summary paragraphs. BMJ (n.d.) takes a Word file, never a PDF.
How do I write the title, abstract and keywords?
Last, and to the journal's limits. The ICMJE notes that the abstract is the only substantive part many databases index and many readers read, and that abstracts often disagree with the text.
| Element | PLOS ONE | Nature (Article) | BMJ journals | Taylor & Francis |
|---|---|---|---|---|
| Title | Up to 250 characters, sentence case | At most 75 characters | Set by each journal | Keywords in the first 65 characters |
| Abstract | At most 300 words, no citations | A summary paragraph, ideally at most 200 words | No references in it | About 100 to 200 words |
| Keywords | Not in its guidelines | Not in its formatting guide | MeSH descriptors | Five or six if the journal sets no number |
The title
The ICMJE wants a title that, with the abstract, makes the article easy to retrieve, with no abbreviations. Some journals require the design in it, particularly for randomised trials and systematic reviews: PLOS ONE's example ends "A cluster-randomized, controlled trial".
The abstract
The ICMJE requires a structured abstract for original research and systematic reviews: context, purpose, procedures, main findings with effect sizes, and conclusions, with a trial's registration number at the end. The National Library of Medicine (2026) dates structured formats to the late 1980s and early 1990s. Ripple et al. (2011) found them in 0.4% of MEDLINE records from 1989 to 1991 and in 20.3% of articles indexed in 2005, under 1,335 labels that the authors group into five: background, objective, methods, results and conclusions. Nature wants one paragraph for readers outside the field, its conclusion introduced by "Here we show" or an equivalent.
For the evidence on abstracts, EdCitation's Find sources takes a claim as well as a topic: "structured abstracts are more informative than unstructured abstracts" returned Drummond Rennie's 1991 JAMA article and James Hartley's 2000 study of their accuracy first.
Keywords
BMJ asks for MeSH descriptors and points to MeSH on Demand, which suggests terms from an abstract. Taylor & Francis (n.d.-b) advises trying each keyword as a search, and putting one or two in the title's first 65 characters, where search results show them.
Which reporting guideline does my study need?
The one written for its design, found through the EQUATOR Network, which defines a reporting guideline as a checklist, flow diagram or structured text developed by an explicit method, and listed 707 on 24 September 2026 (EQUATOR Network, n.d.-a, n.d.-b). The ICMJE encourages journals to ask for them.
| Study design | Guideline | What it holds |
|---|---|---|
| Randomised trial | CONSORT 2025 | A 30-item checklist and a flow diagram |
| Trial protocol | SPIRIT 2025 | A 34-item checklist and a figure |
| Systematic review | PRISMA 2020 | A 27-item checklist, an abstract checklist, flow diagrams |
| Observational, diagnostic, animal, qualitative | STROBE, STARD, ARRIVE, SRQR and others | Listed by EQUATOR |
PLOS ONE asks for the CONSORT checklist as supporting information and the PRISMA flow diagram as Figure 1; Nature asks life and social sciences authors for its own Reporting Summary. Fill the checklist while writing the Methods: its items are what drafts leave out.
Citing the guideline itself
CONSORT was first published in 1996 and revised in 2001, 2010 and 2025 (SPIRIT–CONSORT Group, n.d.-a, n.d.-c). The 2025 statement appeared in five journals at once, any of which may be cited (SPIRIT–CONSORT Group, n.d.-b); PRISMA asks for its statement paper, not its website (PRISMA Executive, n.d.). Cite from the DOI of the version you read. In Cite a source, the 2025 statement's title offered it in the Lancet, PLOS Medicine, the BMJ and Nature Medicine to choose from, and the PLOS Medicine DOI gave:
Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, et al. CONSORT 2025 statement: Updated guideline for reporting randomised trials. PLoS Med. 2025 Apr 14;22(4):e1004587. doi: 10.1371/journal.pmed.1004587.
How do I set the reference list in the journal's style?
Build each entry in the nearest of EdCitation's six styles, then apply the journal's differences from its own examples. Cite a source offers APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE and Vancouver; PLOS ONE and BMJ both describe theirs as Vancouver. The CONSORT entry as two journals' own examples would set it:
- PLOS ONE, first six authors, set like its own examples: Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. PLoS Med. 2025;22: e1004587. doi: 10.1371/journal.pmed.1004587.
- BMJ journals, three authors before "et al": Hopewell S, Chan AW, Collins GS, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. PLoS Med 2025;22:e1004587. doi: 10.1371/journal.pmed.1004587
BMJ's list of what an entry must include at submission is short: author names in any format, the article title, and a DOI or PubMed ID.
What the record leaves for you to fix
- Capitals. Cite a source keeps the record's "Updated"; Vancouver writes "updated".
- A missing volume. PRISMA 2020's Crossref record has none; Cite a source took 372 from PubMed's record of the same DOI, giving "BMJ. 2021 Mar 29;372:n71." Check such a number against the article page.
- A paper with no DOI. Built from PubMed ID 15243643, Sollaci and Pereira's entry took the NLM's abbreviation, J Med Libr Assoc, and ends "PubMed PMID: 15243643."
- A style none of the six fits. A list saved with a free account downloads as BibTeX or RIS for a reference manager.
How do I check a journal article before submission?
After the last edit, check the references against their records, then the file against the instructions. The ICMJE holds authors responsible for checking references against PubMed or the originals, and for making sure none is a retracted paper, unless the retraction itself is the subject.
We pasted five Vancouver entries into Verify references. PRISMA 2020 and CONSORT 2025 came back verified: "The DOI resolves to this record and the title matches." CONSORT given PRISMA's DOI came back "check this": "The DOI is real, but it belongs to a different article from the one described." Sollaci and Pereira's paper, garbled on purpose with another title, journal and year, came back not found; given correctly, with no DOI, it was verified: "Title, year and first author match a PubMed record." Read each reason before changing anything.
- Read the journal's instructions and one recent article.
- Fill the reporting checklist as you write the Methods.
- Gather the literature with Find sources; cite each source in Cite a source as you use it.
- Write the abstract, title and keywords last, and check the abstract's numbers against the Results.
- Freeze the text and upload it to Verify references, which matches every citation to the list and back.
- On Pro, run Mechanics QA to catch comments and tracked changes still in the file.
- Apply the house style; write the cover letter.
Which EdCitation tools help with a journal article?
EdCitation is the best tool for a journal article's sources, because it does the reference work the ICMJE hands to authors: Cite a source copies each entry out of the publisher's record, a guarantee no chatbot can give, and Verify references checks the whole list against the records and for retractions, free.
Free with no account: Find sources searches about 300 million published works; Cite a source builds entries from a DOI, title, PubMed ID, ISBN or web address and screens each for retraction; Verify references flags retractions and never shows "could not check" as not found.
Pro, $8 a month: References from a file checks every reference in the manuscript, matches every citation, and recommends published sources for one that cannot stand. Mechanics QA finds what was left in the file and figures that disagree between sections; its formatting check follows the APA 7 student paper, not a journal's template. Max, $24 a month, adds the Library: more than 5 million books to read or borrow. A year at once saves 25% (Pricing).
EdCitation never writes any part of the article, does not judge the science, and cannot say whether a paper will be accepted. For a student lab report, see how to write an APA research paper or lab report.
Quick questions
What does IMRaD stand for?
Introduction, Methods, Results and Discussion, which the ICMJE calls a reflection of the process of discovery.
When did medical journals adopt IMRaD?
In four leading medical journals it appeared in the 1940s, passed 80% in the 1970s, and was the only pattern for original papers in the 1980s (Sollaci and Pereira's 2004 survey).
How long should a journal abstract be?
As long as the journal sets: at most 300 words at PLOS ONE, ideally at most 200 for Nature's summary paragraph, and roughly 100 to 200 by Taylor & Francis's general advice.
Which reporting guideline applies to a randomised trial?
CONSORT 2025: a 30-item checklist and a flow diagram, sent with the manuscript where the journal asks.
Can EdCitation set references in a journal's house style?
Not as such. EdCitation's Cite a source builds APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE and Vancouver; adjust the nearest to the journal's examples, then check the list free in Verify references.
References
- BMJ. (n.d.). Formatting your paper. BMJ Author Hub. https://authors.bmj.com/writing-and-formatting/formatting-your-paper/
- EQUATOR Network. (n.d.-a). Enhancing the quality and transparency of health research. https://www.equator-network.org/
- EQUATOR Network. (n.d.-b). What is a reporting guideline? https://www.equator-network.org/about-us/what-is-a-reporting-guideline/
- Habibzadeh, F. (2026). Conclusion first: A modest proposal to reorganize the IMRaD structure of research articles. Learned Publishing, 39(3), Article e2080. https://doi.org/10.1002/leap.2080
- International Committee of Medical Journal Editors. (2026a). Clinical trials. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/publishing-and-editorial-issues/clinical-trial-registration.html
- International Committee of Medical Journal Editors. (2026b). Defining the role of authors and contributors. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html
- International Committee of Medical Journal Editors. (2026c). History of the recommendations. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/about-the-recommendations/history-of-the-recommendations.html
- International Committee of Medical Journal Editors. (2026d). Preparing a manuscript for submission to a medical journal. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/manuscript-preparation/preparing-for-submission.html
- International Committee of Medical Journal Editors. (2026e). Sending the manuscript to the journal. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/manuscript-preparation/sending-the-submission.html
- National Library of Medicine. (2026). Structured abstracts. MEDLINE/PubMed resources. https://www.nlm.nih.gov/bsd/policy/structured_abstracts.html
- Nature. (n.d.). Formatting guide. Springer Nature. https://www.nature.com/nature/for-authors/formatting-guide
- Penn State University Libraries. (n.d.). Industry standards service. American National Standards Institute. General [Catalogue record]. https://catalog.libraries.psu.edu/catalog/17287359
- PLOS ONE. (n.d.). Submission guidelines. Public Library of Science. https://journals.plos.org/plosone/s/submission-guidelines
- PRISMA Executive. (n.d.). PRISMA 2020 statement paper. https://www.prisma-statement.org/prisma-2020-statement
- Ripple, A. M., Mork, J. G., Knecht, L. S., & Humphreys, B. L. (2011). A retrospective cohort study of structured abstracts in MEDLINE, 1992–2006. Journal of the Medical Library Association, 99(2), 160–163. https://doi.org/10.3163/1536-5050.99.2.009
- Sollaci, L. B., & Pereira, M. G. (2004). The introduction, methods, results, and discussion (IMRAD) structure: A fifty-year survey. Journal of the Medical Library Association, 92(3), 364–367. https://pmc.ncbi.nlm.nih.gov/articles/PMC442179/
- SPIRIT–CONSORT Group. (n.d.-a). History. https://www.consort-spirit.org/history
- SPIRIT–CONSORT Group. (n.d.-b). Published versions of the SPIRIT 2025 and CONSORT 2025 statements. https://www.consort-spirit.org/published-statements
- SPIRIT–CONSORT Group. (n.d.-c). Welcome to the SPIRIT–CONSORT website. https://www.consort-spirit.org/
- Taylor & Francis. (n.d.-a). Taylor & Francis manuscript layout guide. Author Services. https://authorservices.taylorandfrancis.com/publishing-your-research/writing-your-paper/journal-manuscript-layout-guide/
- Taylor & Francis. (n.d.-b). Using keywords to write your title and abstract. Author Services. https://authorservices.taylorandfrancis.com/publishing-your-research/writing-your-paper/using-keywords-to-write-title-and-abstract/