Live Chat
An article summary restates an article's main claim and the evidence behind it, in your own words, at the length your assignment sets, with the author credited and your own opinion left out. Most guides give you that definition. What guides rarely show is the hard part: deciding which of a real research article's several hundred facts survive the cut, and why.
So this guide works one real study from start to finish. It is a 2019 open-access paper on sleep and grades, about 3,600 words of main text, and we summarize it three times — in one sentence, in one paragraph and in about 250 words — then trace the paragraph version back to the article sentence by sentence and list what we left out. The article is free to read and licensed for reuse, so you can open it alongside this page and check each decision against the original.
- Set the length from your assignment, not from a percentage. Published rules put the ceiling on summary length anywhere from a tenth of the original to half of it. On this 3,640-word study, that is about 360 words under one rule and about 1,820 under another.
- Separate what the authors found from what they cite. An introduction is full of other people's results, and a summary that credits them to the authors misreports the article.
- Keep the authors' verbs. "Associated with" and "leads to" are different claims, and this study's own news release put the stronger one in its headline.
- Do not paraphrase the abstract. It is the authors' condensation of their own work, and this one mentions its gender result without stating it.
- Write the shortest version first. One sentence forces you to find the claim; every longer version adds only what the extra words can buy.
What an article summary is, and what it is not
A summary condenses a whole text for a reader who has not read it. It follows the author's claims, not yours, and it is much shorter than the original. Three neighboring tasks get confused with it, and each one changes what you are graded on.
| Task | What it condenses | The difference that matters |
|---|---|---|
| Summary | A whole article, chapter or study | Your words, the author's ideas, no evaluation |
| Paraphrase | One passage | Roughly the same length as the passage, reworded |
| Abstract | A whole work, usually attached to that work | In a course, the abstract you attach is for your own paper; some courses also assign one for a published article |
| Response, critique or annotation | A summary plus your judgment | Evaluation is required, and it comes after the summary |
The labels are less tidy outside a course. PLOS Computational Biology asks the authors of every research article for "a 150-200 word non-technical summary of the work," and the Journal of Consumer Research asks authors to summarize their own accepted paper in "about 600 words." In the other direction, the University of North Carolina Writing Center's handout on abstracts has a section headed "If you are abstracting someone else's writing." So the reliable distinction is the one your assignment uses: the article summary you are assigned condenses someone else's article for a reader who has not read it, and the abstract you attach to a class paper condenses your own paper. Some courses also assign an abstract of a published article as practice; that is still an abstract, written to an abstract's conventions.
If your assignment adds evaluation to the summary, as an annotated bibliography entry does, the summary is still the first move and still follows every rule below.
The article we will summarize
Kana Okano, Jakub Kaczmarzyk, Neha Dave, John Gabrieli and Jeffrey Grossman gave wearable activity trackers to 100 students in an introductory chemistry course at MIT and compared a semester of sleep data with the students' quiz and midterm scores. The paper appeared in npj Science of Learning on October 1, 2019, under the title "Sleep quality, duration, and consistency are associated with better academic performance in college students."
We chose it for four reasons. It is open access under a Creative Commons Attribution 4.0 license, so anyone can read the whole thing. It is about college students. It has the shape of a typical empirical research article — an abstract, then introduction, results, discussion and methods. And it has an abstract, which lets us test a tempting shortcut: paraphrasing the abstract instead of the article.
The word counts are ours, made the way a word processor counts: words separated by spaces, with the citation numbers and figure caption left out.
Step 1: Read for the claim before you read for detail
Do not start at the first line and take notes on everything. Read in this order: the abstract, the first paragraph of the discussion, the headings in the results, and only then the whole article. By the time you reach the methods, you should be able to write the claim in one sentence.
For this article, the discussion opens with it: "This study found that longer sleep duration, better sleep quality, and greater sleep consistency were associated with better academic performance." That sentence is your anchor. Everything you keep later has to support it, size it or limit it.
On the full read, mark each sentence as one of three things: a finding, a piece of method or a limit. Anything that is none of those is a candidate for the cut list.
If your article argues a position instead of reporting a study, the steps still hold. The claim is the author's thesis, the findings are the reasons and evidence offered for it, and the limits are the objections the author concedes or answers. Keep the concessions. A summary that drops them makes the author sound more certain than the article does.
Step 2: Separate what the authors found from what they cite
This is an easy place for a careful summary to go wrong. An introduction reviews earlier research, so it is full of results that belong to someone else.
This one is no exception. Its introduction reports that "the cognitive performance of an individual who has been awake for 17 h is equivalent to that exhibited by one who has a blood alcohol concentration of 0.05%." That is a striking line, and it is not Okano and colleagues' finding. It ends with a citation number, which marks it as earlier work. Two paragraphs later, the introduction describes a pair of earlier studies that offered students extra credit for averaging eight or more hours of sleep; in the psychology course, the students who did so scored better on the final exam. That is not this study either.
A summary that says "Okano et al. found that staying awake for 17 hours impairs you like alcohol" credits five authors with someone else's result. The rule that prevents it is simple: a sentence that ends in someone else's citation is background. The authors' own findings live in the results, and their interpretation of those findings lives in the discussion. If you genuinely need a cited result, cite its original source, or follow your style guide's rule for citing a source you found inside another one.
On October 8, 2026, we tried to open every text result (videos and social posts aside) on five Google results pages for this topic and read the 45 pages that teach summary writing or set a summary assignment. None of the 45 warns against crediting an article's cited studies to its authors. Two come close without naming the mistake. Grammarly's guide to summarizing a research paper says you might need to cite "the name of an author whose work the original cites," as a way to avoid plagiarism, and a Trent University handout's sample summary separates the scholars an article cites from what its authors did. Another university handout tells students that the first section of a summary "should include all the important background information and context." That is fair advice, provided the summary also says whose background it is.
Step 3: Set the length from the assignment, not a percentage
Nine of those 45 pages give summary length as a share of the original. They disagree, and on a full research article the disagreement is not small. Here are five of their rules, applied to this article:
| Length rule, as published | Where it appears | Applied to this 3,640-word article |
|---|---|---|
| "usually no more than 10 percent" (for longer papers) | Grammarly, guide to summarizing a research paper | no more than about 360 words |
| "about 10-15%" | Documind, guide to summarizing journal articles | about 360 to 550 words |
| "about one-third to one-fourth" | Jenni, summary guide | about 910 to 1,210 words |
| "about one third of the length of the original article" | Summary-writing handout hosted on IHMC's CmapTools server | about 1,210 words |
| "up to 50% of the original" | EAP Foundation, guide to summary writing | anything up to about 1,820 words |
Documind adds that your professor's specific guidelines "trump any rule of thumb," which is the right instinct. The trouble is the fallback. Fixed word counts spread just as widely. Editage's guide says a strong summary of a full research paper "rarely exceeds 200–300 words." A University of Texas at Austin chemical engineering course sets 300-350 words. wikiHow allows "no more than 500-1000 words, for most academic articles."
The percentage rules work on short texts, and the examples printed beside them are short. Five of the 45 pages print a source text next to a shorter version written from it, and the longest of those sources runs about 600 words. San José State University's writing center says of its own sample, "Remember that this sample article is short." A fixed fraction of 600 words is a usable target. A fixed fraction of a 3,600-word study can be longer than the whole assignment.
So use the number your instructor gives you. If there is no number, ask. If you cannot ask, choose the length by the job the summary has to do. The University of Toronto's OISE Academic Skills Hub makes the same point: "The appropriate length of a summary and level of details depend on the task you are working on." In practice, that means one sentence when the summary introduces a source inside your own argument, a paragraph for an annotation or a literature review entry, and a longer version only when the summary is the whole assignment.
Step 4: Write the shortest version first, then expand
Starting with one sentence forces you to commit to the claim. Each longer version then adds only what its extra words can pay for. One of the 45 pages does show a single source at three lengths. The OISE guide quoted above takes Jonathan Swift's "A Modest Proposal," which it describes as a 3,400-word satirical essay, and gives summaries of 11, 94 and 230 words (our counts). It judges the shortest under-summarized, the middle one adequate and the longest over-summarized, and notes that the shortest and the longest could each suit a different task. None of the 45 does this with a research article, or shows where each sentence of a summary came from, so here is the same study three ways. The word counts are whitespace counts, the way your instructor's word processor will count them.
Using wearable trackers on 88 students in an MIT chemistry course, Okano et al. (2019) found that longer, better, and more consistent sleep across a semester was associated with higher quiz and exam scores, but sleep on the night before a test was not.
In a study published in npj Science of Learning, Okano et al. (2019) gave Fitbit activity trackers to 100 volunteers in an introductory chemistry course at MIT and analyzed a semester of sleep data, along with quiz and midterm scores, for the 88 who completed the study. Students who slept longer, slept better according to the device's quality score, and had less day-to-day variation in how long they slept tended to score higher; together, the three measures accounted for about a quarter of the variation in scores. Sleep on the night before a quiz or midterm showed no significant relationship with that score, while average sleep over the month before each midterm did. Female students outscored male students, but the gap was no longer significant once sleep quality was taken into account. The authors note that showing cause and effect would require randomized trials and that their sample came from a single course at one university.
Earlier studies linking sleep to grades, Okano et al. (2019) note, mostly relied on students' own reports of how they slept. To measure sleep objectively instead, they gave Fitbit activity trackers to 100 volunteers in an introductory chemistry course at MIT and, for the 88 who completed the semester, compared the devices' sleep records with scores on eight quizzes and three midterms.
On average, the students went to bed at 1:54 a.m. and slept a little over seven hours a day. Those who went to bed and woke up earlier tended to score higher, as did those who slept longer, slept better according to the device's quality score, and had less day-to-day variation in how long they slept. Together, duration, quality and consistency accounted for about a quarter of the variation in scores. The results also differed by time window: sleep on the night before a quiz or midterm showed no significant relationship with that score, whereas average sleep over the month before each midterm did, which the authors take to suggest that sleep matters most while material is being learned. Female students outscored male students, but the difference was no longer significant once sleep quality was taken into account.
They list several limitations. The quality score comes from a proprietary algorithm that, they note, has no published validation. The relationship may also depend on other factors, such as stress and motivation, and establishing cause and effect would require randomized trials. And the participants were mostly first-year students in one MIT course, so the results may not generalize.
Look at what each step up bought. The paragraph adds that the 88 were what remained of 100 volunteers, the one number that sizes the effect, the gender result and the limits. The 250-word version adds the gap the study was designed to fill, the bedtime finding, the authors' interpretation of the timing result and the reason to be careful with the device's quality score. Nothing in a longer version contradicts a shorter one. That is the test of a well-built set.
Where each sentence came from, and what we cut
One of the three markers of a strong summary in Editage's guide is "Every sentence traces back to something in the original paper." Here is that trace for the one-paragraph version: the selection process made visible, the part a finished summary hides.
| Sentence in the summary | Where it comes from | The decision behind it |
|---|---|---|
| 1. Who, where and how many | Abstract; Methods (Participants, Assessments) | Kept 100 and 88. Cut the 370 students enrolled, the first-100 selection rule and the device model. |
| 2. Longer, better, more consistent sleep went with higher scores | Results, "Sleep duration, quality, and consistency" | Wrote "less day-to-day variation in how long they slept" because the paper defines inconsistency as the standard deviation of each student's daily sleep duration, not as a regular bedtime. |
| 3. About a quarter of the variation | Results, variance section; last paragraph of the Discussion | Kept one number that sizes the finding. Cut the regression equation and the 24.44% figure. |
| 4. Night before versus the month before | Results, "Timing of sleep" | Kept "no significant relationship," the strength the data support, rather than "sleep the night before does not matter." |
| 5. The gender result | Results, "Gender differences" | Kept the score gap and what happened when sleep quality was controlled. Cut the male-only and female-only correlations. |
| 6. The limits | Discussion, limitations paragraph | Kept cause and generalizability. Moved the device-validity limit up to the 250-word version. |
And here is what the article contains that none of the three versions mentions:
- The 17-hour and blood-alcohol comparison, and the extra-credit studies. Both are earlier research the introduction cites, not this study's results.
- How the tracker estimates sleep. Useful if your course is about measurement; otherwise it is method detail.
- The week before each quiz. The authors also tested sleep in the week before each quiz. For sleep duration they call the results "largely consistent" with the midterms (significant on 3 of 8 quizzes, marginal on 3 more); for sleep quality, 1 quiz was significant. The month-before-midterm result held for every midterm, so we kept that one.
- Every correlation coefficient and p-value. Most summary assignments do not ask for test statistics. If yours does, the results section has them.
- The 22 students who also took an exercise class. That group belonged to a separate research question, and the authors report that they performed similarly to the others.
- Why the final exam and the ninth quiz were left out of the score. MIT grades first-semester freshmen on a pass/no record basis, and the authors worried about motivation late in the term. That matters if you are critiquing the method. A summary can let it go.
Step 5: Keep the authors' verbs
The paper's title says sleep is "associated with" better performance. MIT's news office announced the same study under the headline "Better sleep habits lead to better college grades" and called it "an experiment." The sleep findings did not come from an experiment: the researchers measured how students slept; they did not assign anyone more or less sleep.
Further down, the release does concede that "this correlation can't absolutely prove that sleep was the determining factor in the students' performance, as opposed to some other influence that might have affected both sleep and grades." It also quotes Robert Stickgold, a Harvard Medical School sleep researcher who was not connected with the study. He has "no doubt" that shorter, more irregular sleep will hurt a student's grades, but he adds that "correlation is not the same as causation": doing poorly in class might itself disrupt sleep, or a third factor such as ADHD might affect both. A reader who stops at the headline meets none of that. "Associated with" and "leads to" are different claims, and only the first is supported by this design.
The article itself reaches for causal language in its discussion: "The outcome that better 'content-relevant sleep' leads to improved performance is supported by previous controlled studies on the role of sleep in memory consolidation." Read the end of that sentence: the causal reading leans on other researchers' experiments, which is Step 2 again. The limitations paragraph sets the standard for this study's own data: "Establishing a causal relation between sleep and academic performance will require experimental manipulations in randomized controlled trials." Summarize the finding at that strength, and if you mention the causal interpretation, attribute it.
| What the article supports | What a careless summary says | Why the difference matters |
|---|---|---|
| Better sleep was associated with higher scores | Better sleep leads to higher grades | The study observed students; it did not change their sleep |
| No significant relationship with sleep the night before | Sleep the night before does not matter | Not detected in 88 students is a weaker claim than "does not matter" |
| The gender gap was not significant once sleep quality was controlled | Women get better grades because they sleep better | Controlling for a factor in a statistical model does not show that it causes the gap |
Several of the 45 pages warn in general terms against overstating an author's findings, and one says a summary should reflect the authors' "degree of certainty." None of them names this particular slip, from association to cause, which is the one a research article invites.
Step 6: Check it against the article, not the abstract
Two guides seem to disagree about the abstract. One political science instructor's guide wants the opening paragraph of an article summary, about 150-200 words, to "resemble the abstract of the paper itself," in your own words, and then asks for sections on the literature, on the design and results "in some detail," and on the conclusions. wikiHow warns that "an article summary that looks just like the abstract is a poor summary." They agree more than they seem to: the abstract can shape your opening, but it cannot stand in for the article. The worked example shows why.
The abstract is the authors' own condensation of their work, written so a reader can decide whether to read the paper. This one is 162 words, and a summary built from it alone would be missing three things:
- The gender result. The abstract's last sentence is "Gender differences are discussed." It tells you a result exists without telling you what it is.
- The bedtime finding. Earlier bedtimes and earlier wake-up times went with higher scores. The abstract does not mention bedtimes or wake-up times.
- Every limitation. The unvalidated quality score, the single course and the causation caveat all live in the discussion.
Use the abstract as a map for your first read. Build the summary from the article, and check the finished version against the article too.
Naming and citing the article inside your summary
Name the authors and the year in your first sentence, and give the full reference at the end or on your reference page. In APA 7, a work by three or more authors is cited with the first author's name and "et al." from the very first citation, which is why all three versions above say "Okano et al. (2019)." This article uses an article number instead of page numbers, and APA 7 writes the word "Article" before that number:
Okano, K., Kaczmarzyk, J. R., Dave, N., Gabrieli, J. D. E., & Grossman, J. C. (2019). Sleep quality, duration, and consistency are associated with better academic performance in college students. npj Science of Learning, 4(1), Article 16. https://doi.org/10.1038/s41539-019-0055-z
MLA 9 formats the same source differently, and our APA 7 and MLA 9 comparison sets the two side by side. Whichever style you use, remind the reader every few sentences whose ideas these are: "the authors report," "they note," "the study found." Neutral reporting verbs keep your voice out of a text that is supposed to be theirs.
Before you submit: a seven-point check
- The first sentence names the authors and the year and states what they found or argued.
- Every finding in your summary comes from the results or discussion, not from a citation in the introduction.
- Your verbs are no stronger than the article's. "Associated with" stays "associated with."
- The numbers you kept size the main finding. Test statistics stay out unless the assignment asks for them.
- The length matches the assigned word count, counted the way your instructor will count it.
- No sentence states your opinion. If the assignment wants a response, it comes after the summary and is clearly marked.
- You checked the summary against the article, not against the abstract.
One more check if you drafted with an AI summarizer: whether that was allowed is your instructor's call, so read the syllabus first. Our guide on using AI for essays covers where those policies usually draw the line. Either way, run Steps 2 and 5 on every sentence, because a fluent summary can still credit the authors with someone else's finding or upgrade their verb.
Article summaries rarely come alone. They feed reading responses, annotated bibliographies and literature reviews, often several in the same week. Where your course permits outside help, our coursework help gives you a draft to review, check against the article and learn from, and what you submit is your own. Where your syllabus rules that out, your instructor and your campus writing center are the right help. Either way, the method above is the one to hold any summary to.
