Contents
On this page
Academy · Original research article · Writing an original article
Writing an Original Article: An Overview of the IMRAD Structure for Beginners
Each IMRAD section addresses one question, and its content is specified rather than a matter of preference. Most structural problems arise when a boundary between sections is crossed.
An original article reports research conducted by the authors and presents data not previously published. Almost every journal expects it to follow the IMRAD structure (Introduction, Methods, Results, Discussion), and the content of each section is specified rather than left to preference.
Definition of an original article
An original article is the primary form of scientific publication: a report of work the authors did, with data that have not appeared before. That is what distinguishes it from a review, which synthesises other people’s findings, and from a case report, which describes one patient or a small series. If you want the comparison across article types set out properly, our guide to abstract formats covers how each type is summarised.
The reason the structure is worth taking seriously is that it is not decoration. Each section answers one question, and a reader — including a reviewer, and including the editor deciding whether to send your paper out at all — navigates by it. Mateu Arrom and colleagues' review of the form treats the section-by-section requirements as the substance of what makes a paper publishable4.
This paper is sometimes cited as “Arrom, L. Mateu, et al.” The surname is Mateu Arrom, so the correct short form is Mateu Arrom L et al. A search for “Arrom LM” will not retrieve it.
The IMRAD structure and its adoption as a standard
IMRAD is a convention, and a more recent one than its dominance suggests. Sollaci and Pereira surveyed fifty years of articles in four major journals and found the structure “recommended since the beginning of the twentieth century”2 and “was already considered the ideal outline for scientific writing in the first quarter of the 20th century”2 — but that in practice it “began to be used in the 1940s”2 and was “adopted as a majority only in the 1970s”2
The years at which it became the only pattern in each journal they studied2:
Table 1. The year IMRAD became the only structure used for original articles
| Journal | Year |
|---|---|
| New England Journal of Medicine | 1975 |
| BMJ | 1980 |
| JAMA | 1985 |
| The Lancet | 1985 |
By the end of their survey period the structure “reached over 80%”2 of articles. Knowing this is useful for one practical reason: when a journal’s instructions depart from IMRAD — combining results and discussion, or asking for a structured introduction — that is a house rule you follow, not an error on their part.
Content of each section
The ICMJE recommendations set out what each section must contain. Reading them as a pair of columns — what belongs, and what does not — is more useful than reading them as prose, because nearly every structural problem in a submitted manuscript is something in the wrong column1.
Table 2. The four sections, per ICMJE
| Section | What belongs here | What does not |
|---|---|---|
| Introduction | The context, the nature and significance of the problem, and the specific purpose, objective or hypothesis. | No results and no conclusions from the work being reported, and only directly pertinent references. |
| Methods | Sufficient detail to allow others with access to the data to reproduce the results, including a statement that an ethics committee approved the research. | Only what was known when the protocol was written. Any information obtained during the study belongs in the Results. |
| Results | A logical sequence, with the most important findings first, data for all primary and secondary outcomes, and derived values given with the absolute numbers from which they were calculated. | No repetition in the text of all the information already presented in the tables and figures. |
| Discussion | A summary of the main findings, possible explanations, the placement of the work in the context of the wider evidence, and a statement of limitations and implications. | No repetition of data given elsewhere, and no claim of priority or allusion to unfinished work. |
Two of those deserve expanding. In the methods, ICMJE asks for enough detail that others with access to the data could reproduce the results, and that is the test to write against: not “have I described what I did” but “could a competent stranger do it again from this”. In the results, the instruction not to repeat the tables in the text is the one most often ignored — “Do not repeat all the data in the tables or figures in the text.”1 The text should pick out what matters and point at the table for the rest.
Three section boundaries commonly crossed by authors
If the section structure does one thing, it is to keep four kinds of content apart. These are the three boundaries that get crossed, with ICMJE’s own wording on each.
Table 3. The section boundaries, and what ICMJE says
| Boundary | The rule | In ICMJE’s words |
|---|---|---|
| Introduction / Results | The introduction should not contain the findings. | “Cite only directly pertinent references, and do not include data or conclusions from the work being reported.” |
| Methods / Results | The methods describe what was planned; what emerged during the study belongs in the Results. | “all information obtained during the study belongs in the Results section.” |
| Results / Discussion | The discussion interprets the findings rather than restating them. | “Do not repeat in detail data or other information given in other parts of the manuscript” |
The middle one is the subtle one. The methods section describes the study as designed and conducted; anything the study turned up is a finding. A protocol deviation, a recruitment shortfall, an assay that had to be changed partway — these feel like methods, because they are about how the work was done, but ICMJE places them in results1. Getting this wrong hides exactly the information a careful reader is looking for.
One more ICMJE instruction worth naming, because it is about tone rather than placement: in the discussion, “Avoid claiming priority or alluding to work that has not been completed.”1 Claims of priority and allusions to work in progress read as insecurity, and they date badly.
Determination of article length
There is no general answer, and any page that gives you one is inventing it. Two real examples, from journals that publish their limits:
Table 4. Two journals’ word limits for original research
| Journal | Article type | Limit | Notes |
|---|---|---|---|
| Scandinavian Journal of Work, Environment & Health | Original research | 4,000 words of text, introduction to conclusion | 250-word abstract |
| British Journal of Sports Medicine | Original report | 3,000 words maximum | Excludes title page, abstract, tables, acknowledgements, contributions and references |
The first allows “Text (from introduction to conclusion): 4000 words” with an “Abstract: 250 words”5. The second states that “Papers should be a maximum of 3000 words in length.”6, excluding the title page, abstract, tables, acknowledgements, contributions and references6.
The two journals set different limits and apply different rules regarding what counts towards them. This is all that can reliably be stated about length in general: it is set by the journal, it is usually a total rather than a per-section allocation, and the exclusions matter as much as the number. The requirements of the target journal should be checked before drafting, because reducing a manuscript from 4,000 to 3,000 words after completion amounts to a rewrite.
On per-section word counts. Figures for the recommended length of each section circulate widely without attribution. They derive from the guide for beginners by Wenzel and colleagues3:
Table 5. Per-section word counts, as advised by Wenzel and colleagues (2007)
| Section | Suggested length |
|---|---|
| Introduction | about 350 words |
| Methods | about 1,000 words |
| Results | about 350 words |
| Discussion | 1,000 to 1,350 words |
Status of these figures as advice rather than a standard
These figures represent the advice of one group in 2007 rather than a standard. They are useful as an indication of proportion (methods and discussion long, results short), but they do not constitute a limit that will be enforced. The binding limit is that of the journal, which is a total rather than a per-section allocation.
Preparation before drafting
Two things are worth settling before the first draft, because both are expensive to fix afterwards.
The journal. Word limits, reference style, section conventions and reporting-guideline requirements all come from the journal, and all of them shape the draft. Choosing afterwards means rewriting.
Whether you have a paper. A conference abstract is a preliminary report rather than a finished paper, and the distance between the two is wider than most authors expect. (It is worth being precise here rather than saying an abstract “is not a publication”: the ICMJE treats an abstract or poster at a scientific meeting as a preliminary report, which is why presenting does not normally stop a journal considering the full paper afterwards.)
Accuracy of conference-to-publication figures
It is often stated, without a source, that “only around half of the abstracts at scientific conferences are turned into complete works”. The evidence indicates a lower figure. Scherer and colleagues, reviewing 425 reports covering more than 307,000 abstracts, found an overall full publication proportion of 37.3%, with a survival analysis estimating 46.4% published in full by ten years. The figure of “around half” is therefore defensible only as the ten-year estimate, and the headline figure is closer to one third.
Scherer and colleagues reviewed 425 reports covering 307,028 abstracts and found an “overall full publication proportion of 37.3%”7 — 95% CI 35.3% to 39.3% — rising to an estimated 46.4% by ten years7. Nearly two-thirds of work presented at conferences never appears in full. The practical reading is not discouraging but clarifying: the step from abstract to article is the step most research does not survive, and it is worth planning for rather than assuming.
Common deficiencies in original articles
The recurring problems are structural rather than stylistic, and most of them are failures of the boundaries in Table 3: results previewed in the introduction, findings buried in the methods, the results section narrating tables line by line, the discussion restating the results instead of interpreting them, and a conclusion that goes further than the data support41.
Two more, on the statistics. ICMJE asks authors to “Avoid relying solely on statistical hypothesis testing, such as P values.”1 and to give derivatives alongside the absolute numbers they came from — a percentage without its denominator is not a result. If you want that argument in full, our pages on the null hypothesis and on significance versus effect size set out what a p-value does and does not tell a reader.
A recommended order of drafting
Practical application
- The target journal should be selected first, and its instructions to authors read before drafting begins. The word limit, the elements it excludes and the reference style all shape the draft.
- The methods should be written first, as they are the section most constrained by what actually occurred and the least dependent on other sections.
- The results should be written next, as a logical sequence beginning with the most important finding. Absolute numbers should be given alongside percentages, and the text should not narrate the tables.
- Any information obtained during the study should be moved from the methods to the results, where ICMJE places it.
- The discussion should be written from the results: it should summarise the main findings, explain them, relate them to the wider evidence and state the limitations clearly.
- The introduction should be written once its subject is clear, and any result or conclusion that has entered it should be removed.
- The title and abstract should be written last, on the basis of the completed paper.
- The discussion should be checked for claims of priority and for references to unfinished work, which should be removed.
- The word count should be checked against the journal’s limit and its exclusions rather than against a general figure.
Key points
An original article reports the authors’ own previously unpublished research in the IMRAD structure, which became standard only in the 1970s. The International Committee of Medical Journal Editors (ICMJE) specifies the content of each section, and nearly every structural problem involves a crossed boundary: results previewed in the introduction, findings left in the methods, or a discussion that restates rather than interprets. Length is set by the journal, and its exclusions matter as much as the number; the per-section word counts in circulation represent one group’s 2007 advice, not a standard. A conference abstract is not a paper: only about 37% reach full publication.
A report of research carried out by the authors, presenting data not previously published. It is the primary form of scientific publication and is distinguished from a review, which synthesises the work of others, and from a case report, which describes one patient or a small series. The structure expected by almost all journals is IMRAD: Introduction, Methods, Results and Discussion.
Introduction, Methods, Results and Discussion. The structure addresses four questions in sequence: why the study was done, how it was done, what was found and what it means. The fifty-year survey by Sollaci and Pereira found that it was recommended from early in the twentieth century, began to be used in the 1940s and became the majority pattern only in the 1970s; it is therefore a convention, and a relatively recent one, rather than a fixed rule.
The appropriate length is that permitted by the target journal, which varies. The Scandinavian Journal of Work, Environment & Health allows 4,000 words of text from introduction to conclusion with a 250-word abstract; the British Journal of Sports Medicine sets a 3,000-word maximum and excludes the title page, abstract, tables, acknowledgements, contributions and references from the count. The instructions of the target journal should be checked before drafting, because the exclusions matter as much as the number.
No. The ICMJE recommendations state explicitly that an introduction should cite only directly pertinent references and “do not include data or conclusions from the work being reported”. The introduction states the question, and the results answer it. Previewing the answer is one of the three section-boundary errors that reviewers readily identify.
The results report what was found, in a logical sequence with the most important findings first, without repeating in the text all the information already presented in the tables and figures. The discussion interprets the findings: it explains them, relates them to other evidence and states the limitations and implications. ICMJE asks authors not to “repeat in detail data or other information given in other parts of the manuscript” in the discussion.
In the results. This is the most subtle of the section boundaries and the one most often blurred. The methods describe the study as it was designed and conducted; ICMJE states that “all information obtained during the study belongs in the Results section”. A protocol deviation, an unexpected recruitment problem or a changed assay constitutes a finding, not a method.
Most experienced authors do not. The methods and results are the sections most constrained by what occurred and the least dependent on the remainder, so they are usually written first; the discussion follows from the results; the introduction is more readily framed once its subject is clear; and the abstract and title are written last, because they summarise the completed paper.
No, and the gap between the two is larger than is commonly assumed. A Cochrane methodology review of 425 reports covering more than 307,000 abstracts found an overall full publication proportion of 37.3%, with a survival analysis estimating 46.4% published in full by ten years. A conference abstract is a preliminary step, not a completed paper.
References
- International Committee of Medical Journal Editors. Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals. Sections IV.A.3.c to IV.A.3.f. read 15 October 2026. View on ICMJEThe normative statement of what each IMRAD section must contain, and the source for the three boundary rules between sections that authors most often break.
- Sollaci LB, Pereira MG. The introduction, methods, results, and discussion (IMRAD) structure: a fifty-year survey. Journal of the Medical Library Association. 2004;92(3):364–367. read 15 October 2026. View on a hosted full textA survey of when IMRAD became standard. Read from a mirrored full-text PDF rather than the publisher's page, which is on a blocked host.
- Wenzel V, Dünser MW, Lindner KH. Wie schreibe ich eine Originalarbeit? Eine Anleitung für Einsteiger [How do I write an original article? An introduction for beginners]. Der Anaesthesist. 2007;56(8):828–836. doi:10.1007/s00101-007-1151-7 read 15 October 2026. View on Der AnaesthesistA practical guide to turning a finished study into a paper. It is the source of the per-section word counts that frequently circulate without attribution.
- Mateu Arrom L, Huguet J, Errando C, Breda A, Palou J. How to write an original article. Actas Urológicas Españolas (English Edition). 2018;42(9):545–550. doi:10.1016/j.acuroe.2018.02.012 read 15 October 2026. View on Actas Urológicas EspañolasA review of what an original article requires and the errors to avoid, organised by IMRAD section. The first author’s surname is Mateu Arrom; a citation in the form “Arrom LM” will not resolve.
- Scandinavian Journal of Work, Environment & Health. Instructions to authors. read 15 October 2026. View on SJWEHOne of two real journal limits cited on this page in place of a universal figure. 4,000 words of text for original research, with a 250-word abstract.
- British Journal of Sports Medicine. Instructions for authors. read 15 October 2026. View on BJSMThe second real limit, and useful because it states its exclusions: 3,000 words, not counting the title page, abstract, tables, acknowledgements, contributions or references.
- Scherer RW, Meerpohl JJ, Pfeifer N, Schmucker C, Schwarzer G, von Elm E. Full publication of results initially presented in abstracts. Cochrane Database of Systematic Reviews. 2018, Issue 11, Art. No.: MR000005. doi:10.1002/14651858.MR000005.pub4 read 15 October 2026. View on Cochrane LibraryA review of 425 reports covering 307,028 abstracts. Read from the James Lind Library's summary rather than the Cochrane record. The source for correcting the claim that about half of conference abstracts reach full publication.
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica’s medical writers and editors support original research articles from draft to acceptance: structuring the manuscript according to IMRAD and the target journal’s instructions, checking that each section contains what ICMJE requires, editing to the word limit, and preparing the submission package. Revision in response to structural comments from reviewers can be undertaken as a discrete piece of work.