An Empirical Examination of the Text Structure of Original Research Articles in Medical Journals
Original Research Article Sample Work
A meta-research paper that counts paragraphs, not patients
This original research sample is unusual among medical manuscripts because its unit of analysis is the published article itself. It asks a practical question that many first-time authors raise: how long should each part of an IMRAD paper be? Rather than offer opinion, it measures the number of paragraphs in the introduction, methods, results and discussion of original research articles in leading general medical journals, and it extends the count to tables, figures, references and supplementary material.
The audience is researchers, clinicians and editors interested in scientific writing, and the paper is written so that its conclusion can be applied directly: a target paragraph structure for authors who have no journal-specific template to follow.
From Hill's four questions to a 2/7/7/6 rule of thumb
The introduction gives a concise history of the IMRAD format. It notes that IMRAD was proposed around the turn of the twentieth century, became the dominant structure from 1965 and was effectively universal by the 1980s. It links the shift to a 1965 meeting of medical editors at which Hill framed a research article as answers to four questions: why did you start, what did you do, what answer did you get, and what does it mean?
The paper then reviews earlier attempts to quantify section length. Albert's analysis of articles from six journals produced a 2/7/7/6 paragraph structure, with a more lenient 2-3/4-6/4-6/5-8 range in an earlier publication, while Araujo's work on cardiology journals suggested 3/6-9/4-9/10. Re-analysing Albert's data with the DerSimonian and Laird method gives 3/8/7/7. Because these estimates were more than two decades old, the authors hypothesise that the spread of reporting guidelines has lengthened modern articles, particularly their methods sections.
Sampling 450 articles across five journals and three years
The methods section describes a clear and reproducible sampling frame:
| Design element | What the sample reports |
|---|---|
| Journals | The BMJ, JAMA, The Lancet, NEJM and PLOS Medicine |
| Years | 2005, 2010 and 2015 |
| Articles | 30 randomly selected per journal and year, 450 in total |
| Outcomes | Paragraphs per section; numbers of tables, figures and references; presence of supplementary material |
| Analysis | Random-effects meta-analysis (DerSimonian and Laird) and linear mixed models, using R 3.6.2 |
Linear mixed models with journal as a random effect and year as a fixed effect estimate change over time, and a separate model tests whether online-only PLOS Medicine carries more paragraphs than the four print journals. The significance level was set at 0.05 without adjustment for multiple testing, and the authors state this openly.
What the counts revealed about the methods section
Every one of the 450 articles followed IMRAD. Total length rose by 1.08 paragraphs every two years (95% CI 0.70 to 1.46), with the methods section growing fastest at 0.29 paragraphs per year (95% CI 0.19 to 0.39). The typical structure moved from 3/9/9/8 in 2005 to 3/11/10/8 in 2015, and the pooled estimate across all years and journals was 3/10/9/8. PLOS Medicine articles were around 3.5 paragraphs longer than those in print journals and had the most figures and references.
The change in supplementary material was the most striking: no JAMA article in the 2005 sample included it, whereas 29 of 30 did in 2015. The discussion connects the growth in methods text to EQUATOR reporting guidelines such as CONSORT, and translates the findings into a word budget, roughly 100 words per paragraph for a 3,000-word limit spread across 30 paragraphs.
Applying the 3/10/9/8 structure to your own manuscript
The sample is useful in two ways. Its findings give authors a defensible starting plan for a draft, and its own construction shows how a short empirical paper can move from a focused hypothesis to a practical recommendation. The discussion is also candid about limits: it did not examine whether structure affects citation counts, and it records a peer reviewer's suggestion to study media attention and changes in author instructions.
When adapting the approach, authors should check that numbers match across the abstract and the main text. In this draft, for example, the abstract names four journals while the methods list five, the kind of inconsistency that careful scientific editing is designed to catch before submission. The reference list uses an author-date (APA-style) format, which suits a paper that discusses the work of named authors; Pubrica's guide to APA, AMA, MLA and Vancouver styles covers the alternatives. The sample's central message is one the authors repeat in their conclusion: a paragraph template is a fallback, and the target journal's instructions to authors take precedence. Researchers who want help structuring a full study can use Pubrica's physician manuscript writing support.
Questions about this sample
What paragraph structure does this study recommend for an original research article?
Pooling all journals and years, the study recommends three paragraphs for the introduction, ten for the methods, nine for the results and eight for the discussion, about 30 in total. The authors present this as a default for situations where the target journal offers no specific structure, not as a rule that overrides author guidelines.
Why did the methods sections of medical articles grow between 2005 and 2015?
The authors attribute the growth largely to reporting guidelines collected by the EQUATOR network, such as CONSORT and STARD, which ask authors to describe design, participants, outcomes and analysis in more detail. Methods sections grew by an estimated 0.29 paragraphs per year, faster than any other section in the sample.
Does an online-only journal publish longer articles than print journals?
In this sample, yes. PLOS Medicine, which has no page limits, carried articles about 3.5 paragraphs longer than BMJ, JAMA, The Lancet and NEJM, and also had the most figures and references. The authors advise checking whether a target journal is print or online-only when planning length.
How were the data analysed in this meta-research study?
The authors calculated means and standard deviations for each journal and year, pooled them with DerSimonian and Laird random-effects meta-analysis, and used linear mixed models to test change over time and differences between journals. All analyses were run in R 3.6.2, with analysis code provided as supplementary material.
Give yourself the academic edge today
Each order includes
- On-time delivery or your money back
- A fully qualified writer in your subject
- In-depth proofreading by our Quality Control Team
- 100% confidentiality, the work is never re-sold or published
- Standard 7-day amendment period
- A paper written to the standard ordered
- A detailed plagiarism report
- A comprehensive quality report
Enquire Now
