Researchers rarely read a paper straight through on the first go. The fastest way to read an academic paper is in passes: a few minutes to decide whether it matters, a second read for its argument and evidence, and a close third read only for the papers your work stands on.
It is also getting harder. Plavén-Sigray et al. (2017) scored 709,577 abstracts from 123 journals, published between 1881 and 2015, and found that "the readability of science is steadily decreasing". In 1960, 14% of the abstracts scored below zero on the Flesch Reading Ease scale; by 2015, 22% did.
EdCitation publishes this guide, which works through that readability paper in three passes. EdCitation's free Find sources found it, Cite a source built its reference, and Verify references checked the key sources.
How do you read a research paper quickly?
You read a research paper quickly by deciding early whether it deserves more time, and by reading its parts out of order:
- Know why you are reading. The University of York Library (2026) suggests asking what you want from the paper and when you will have read enough.
- Get the right paper, in the right version. Search by topic or by the claim you need to support in EdCitation's Find sources, and check whether you have a preprint or the published version.
- First pass, five to ten minutes: title, abstract, introduction, headings, conclusions, a glance at the references.
- Decide: stop, read background first, or go on.
- Second pass, up to an hour: figures and main points, with notes and a list of references to chase.
- Third pass, only when you must rely on or critique the paper: rebuild the work and test every assumption.
- Cite as you go. Paste the DOI into Cite a source, and run the finished list through Verify references.
What is Keshav's three-pass method?
The three-pass method reads a paper in up to three rounds, each with its own goal, as set out by S. Keshav of the University of Waterloo in a two-page article. Keshav (2007) wrote it because reading papers is a skill "rarely taught", which leaves new graduate students to learn by trial and error. Each pass builds on the last, and you may stop after any of them.
The first pass: a bird's-eye view
The first pass takes about five to ten minutes: the title, abstract and introduction carefully, then only the headings, then the conclusions, and a glance at the references for ones you know. You should then be able to answer Keshav's five Cs:
- Category: what kind of paper is it?
- Context: which papers and theories does it build on?
- Correctness: do the assumptions look valid?
- Contributions: what does it add?
- Clarity: is it well written?
Keshav says one pass is enough for a paper outside your area, and that you may stop there if it does not interest you, you lack the background, or the assumptions look wrong.
The second pass: the content, not the details
The second pass takes up to an hour. Read with more care but skip details such as proofs, and jot down key points. Look hard at graphs: are the axes labelled, and do error bars support the conclusions? Mark unread references worth following. Afterwards you should be able to summarise the paper, with its evidence, to someone else. If it still makes no sense, Keshav offers three choices: set it aside, read background first, or go on to the third pass.
The third pass: re-create the work
The third pass is for full understanding, and for reviewers. The key, in Keshav's words, is to "virtually re-implement" the paper: take the authors' assumptions, rebuild the work, and compare your version with theirs, which exposes both the innovations and the hidden failings. At the end you should be able to reconstruct the paper from memory and name its weak points, missing citations and doubtful techniques.
| Pass | What to read | What to ask | Time (Keshav, 2007) |
|---|---|---|---|
| First | Title, abstract, introduction, headings, conclusions; glance at references | The five Cs: category, context, correctness, contributions, clarity | About five to ten minutes |
| Second | The whole paper except details such as proofs; figures and graphs with care | Can I summarise it, with its evidence? Which references next? | Up to an hour |
| Third | Everything, re-creating the work from the authors' assumptions | Which assumptions are hidden or weak? What is missing? | About four or five hours for beginners, about an hour for experienced readers |
| Before and after | Find sources for the paper and related work; Cite a source for its reference | Is this the published version? Is it retracted? | Keshav gives none |
Keshav also applies the method to a literature survey, starting from three to five recent papers; that search, step by step, is in how to find sources for a literature review fast.
How is an empirical research paper laid out?
Most empirical papers in the sciences and social sciences follow IMRaD: Introduction, Methods, Results and Discussion, with an abstract before and references after. Sollaci and Pereira (2004) sampled the BMJ, JAMA, The Lancet and the New England Journal of Medicine from 1935 to 1985: IMRaD appeared in the 1940s, reached 80% of papers in the 1970s and was the only pattern for original papers in the 1980s. They note that the layout suits readers who browse sections for specific information, which is how the method above reads. The form's history is in how to write a journal article.
Where the layout bends
Not every journal keeps the order. The eLife paper worked through below puts Materials and methods after the Discussion. Use the headings on the page, not the order you expect.
What should you read first, by type of paper?
What you read first depends on the kind of paper:
- Empirical study. Abstract, conclusion or discussion, then method and results, to judge whether the conclusion follows from what was done (University of York Library, 2026).
- Theoretical paper, common in the humanities. Abstract, then the body of the argument and its summary. UC Merced Library (2026) notes that humanities articles may have no abstract, so start with the introduction and headings.
- Review article. UC Berkeley Library (2026) notes that literature reviews rarely present original data, so most have no Methods or Results. Read the abstract, the scope and the headings; for a systematic review, check how studies were searched for and chosen (types of literature review). In a new field, Keshav advises looking for a recent survey first.
- Preprint. Read it as a study, remembering that a preprint has not been through a journal's peer review yet.
What five university libraries recommend
The guides agree on one thing: not front to back first. They differ on what follows the abstract.
| Library guide (last updated) | Order it recommends |
|---|---|
| University of Illinois Urbana-Champaign (7 March 2025) | Abstract, discussion and conclusions, then back to methods or results for more |
| University of California, Merced (1 May 2026) | Abstract; skim the introduction, read the conclusion; figures and tables; then a full read with notes |
| University of Michigan (2 September 2026) | Abstract, conclusion, introduction, results, methods, discussion and limitations, then one read-through; it also teaches ADIRM (abstract, discussion, introduction, results, methods) |
| University of York (2 September 2026) | Title and abstract, conclusion, then method and results (empirical) or body and summary (theoretical) |
| University of California, Berkeley (15 September 2026) | Title, abstract, introduction, discussion; then results with the figure captions, methods, references |
The disagreements are worth knowing. York says the introduction can often be skipped; Berkeley and Keshav read it early. Merced calls methods and results the "dry stuff"; York and Illinois send you to them to test the conclusion. Our view: the methods can wait if you need only the gist, and cannot if you will rely on the finding. Michigan adds that an abstract is too brief to replace the article (University of Michigan Library, 2026).
A worked example: one open-access paper in three passes
The example is Plavén-Sigray et al. (2017), "The readability of scientific texts is decreasing over time", in eLife, open access under a CC BY licence, read in full in September 2026.
First pass
The abstract gives the corpus, the finding and a cause: more general scientific jargon. Each Results heading states a finding as a sentence, so the headings alone tell the story: abstracts have grown harder to read, in most journals and every field studied; abstracts track full texts; and readability falls with more authors and more jargon. The five Cs:
- Category: a measurement study of published text.
- Context: two readability formulas, Flesch Reading Ease and New Dale-Chall, and earlier studies of language in science.
- Correctness: two assumptions to watch: that a formula captures difficulty, and that abstracts stand for whole papers.
- Contributions: the decline, and evidence that jargon explains part of it, while more authors do not explain it away.
- Clarity: readable; the authors report that their article scores 49 on Flesch Reading Ease and its abstract 40.
Second pass
Figure 2 carries the main result. Read its caption first: on Flesch Reading Ease a lower score means harder text, on New Dale-Chall a higher one does. The yearly average Flesch score fell with year (r = -0.93). On the list to chase: Hartley's 1994 study, cited for the finding that rewritten abstracts were better understood.
The note, in our words: abstracts in 123 highly cited journals have become harder to read, partly through rising scientific jargon; readability formulas are rough measures, as the authors say.
Third pass
The journals were highly cited titles indexed in PubMed, so "science" here is mainly biomedicine. The check that abstracts stand for full texts used six young open-access journals outside the main list (r = 0.60 for Flesch scores). The four authors rated the jargon list themselves and say their ratings were not fully independent. The code is public on GitHub, so Keshav's virtual re-implementation can become a real one.
How do you take notes while reading a paper?
Take notes you can cite from without reopening the paper. Add three habits to Keshav's key points:
- Record the full reference first, from the DOI. EdCitation's Cite a source builds it in six styles.
- Write points in your own words, and put exact wording in quotation marks with its page or section. Which to use when is in paraphrase, quote or summarise.
- Keep the paper's claim apart from your judgement of it.
If a note rests on a paper cited inside another, read the original or cite it as a secondary source, as cited a source you did not read explains.
How do you judge a paper's quality in a few minutes?
Judge the paper as you pass through it, with five checks:
- Is it peer reviewed? The eLife paper shows received and accepted dates, 12 April and 4 August 2017. See how to tell if an article is peer reviewed.
- Is it the published version? The same study was posted on bioRxiv on 22 March 2017 with different numbers: 707,452 abstracts from 122 journals. Cite the version of record.
- Do the conclusions match the data? Check the results and the limitations section.
- Who paid, and who gains? Berkeley points to the acknowledgements for conflicts of interest. This paper declares none.
- Is the reference real and not retracted? How to check a reference is real gives the steps; EdCitation does them below.
How does EdCitation help you find, cite and check a paper?
EdCitation is the best tool for the reading work that happens outside the paper, because it looks every reference up in the publisher's record and never writes one. A chatbot composes references; EdCitation checks them against the record. It never writes any part of a student's work, and these three tools are free, with no account.
Find the paper and the work around it
Find sources searches about 300 million published works by topic or by the claim a sentence needs to support. On 26 September 2026 we searched "readability of scientific abstracts decreasing over time", with no filter or sort chosen. It reported 4,682,783 results. The first was the bioRxiv preprint, "The readability of scientific texts is decreasing over time" (2017), type preprint, cited by 4; the second, a 2018 Faculty Opinions recommendation of it. The eLife version was not among the 96 results returned; bioRxiv's record of the preprint names it as published in eLife under DOI 10.7554/eLife.27725. Further down were follow-up studies for forward reading, among them "Readability is decreasing in language and linguistics" (2022, Scientometrics, cited by 34).
Cite it from the DOI
Cite a source builds the reference in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE and Vancouver, and screens for retraction. From 10.7554/eLife.27725 it returned, from Crossref's record, with no retraction notice:
Plavén-Sigray, P., Matheson, G. J., Schiffler, B. C., & Thompson, W. H. (2017). The readability of scientific texts is decreasing over time. eLife, 6, Article e27725. https://doi.org/10.7554/eLife.27725
On the page, eLife and the volume 6 are in italics, as APA 7 requires.
Verify the list
Verify references checks a pasted list or an uploaded paper against the publishers' records, flags retractions, and never shows "could not check" as "not found". We pasted Keshav (2007), Plavén-Sigray et al. (2017) and Sollaci and Pereira (2004). All three came back verified: the first two with "The DOI resolves to this record and the title matches.", the third with "Title, year and first author match a PubMed record."
With Pro, $8 a month, References from a file checks every reference in an uploaded paper and matches each in-text citation to the list; see Pricing.
Quick questions
What are the three passes in Keshav's method?
A first pass of five to ten minutes for the gist and the five Cs, a second of up to an hour for the content and figures, and a third that re-creates the work to expose its assumptions (Keshav, 2007).
How long should it take to read a research paper?
Keshav gives five to ten minutes for the first pass, up to an hour for the second, and four or five hours for a beginner's third, about an hour for an experienced reader.
Should I read the abstract or the conclusion first?
Read the abstract first. Every library guide we read puts it first, and most go to the conclusion or discussion next.
Is reading only the abstract enough to cite a paper?
No. The University of Michigan Library says an abstract is too brief to replace the article, and York warns that an abstract may claim a stronger outcome than the paper reaches.
How do I find papers to read on my topic?
Search EdCitation's free Find sources by topic or by the claim you need to support, give each promising result a first pass, and cite the keepers in Cite a source.
References
- Keshav, S. (2007). How to read a paper. ACM SIGCOMM Computer Communication Review, 37(3), 83–84. https://doi.org/10.1145/1273445.1273458
- Plavén-Sigray, P., Matheson, G. J., Schiffler, B. C., & Thompson, W. H. (2017). The readability of scientific texts is decreasing over time. eLife, 6, Article e27725. https://doi.org/10.7554/eLife.27725
- 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/
- UC Berkeley Library. (2026, September 15). College Writing 161: Writing in the biological sciences: Read scientific articles. https://guides.lib.berkeley.edu/cw161/read
- UC Merced Library. (2026, May 1). How to read a scholarly article: Key reading strategies. https://libguides.ucmerced.edu/reading-scholarly-articles/strategies
- University of Illinois Urbana-Champaign Library. (2025, March 7). How to read a scientific paper: Structure of an article. https://guides.library.illinois.edu/c.php?g=348214&p=2346645
- University of Michigan Library. (2026, September 2). The library research process, step-by-step: Reading scholarly articles. https://guides.lib.umich.edu/c.php?g=1404624&p=10396344
- University of York Library. (2026, September 2). Being critical: Reading academic articles. https://subjectguides.york.ac.uk/critical/articles