How to write an abstract for a scientific article

An abstract is not the introduction to a paper and it is not promotional copy. It should provide a compressed version of the study: why it was done, how it was done, what was found and what follows from the findings. In many databases, the abstract is the main substantive text a reader sees before the full article, so any mismatch between abstract and manuscript is particularly damaging.

Four core parts

For a study that follows IMRaD logic, it is useful to think in four sentences or four short blocks.

Background/Objective: what problem is being addressed and what is the specific objective? Do not spend two paragraphs on general background that every reader in the field already knows.

Methods: who or what was studied, the design, the main methods and the essential sample size.

Results: the primary result, preferably with concrete numbers where appropriate. “Significant differences were found” is weaker than stating the direction and magnitude of the finding.

Conclusion: answer the research question within the limits of the data. Do not make a conclusion broader than the population or design allows.

A structured abstract separates these parts with headings. An unstructured abstract presents the same logic in one paragraph. The journal decides which format is required.

How long should it be?

There is no universal word count. Many journals set limits around 150–300 words, but other ranges exist. Follow the author guidelines. If the limit is 250 words, do not spend 90 on background and only 15 on results.

When shortening, remove generic introductory phrases before removing numbers or limitations. “Problem X is widespread and of great interest” usually contributes less than a precise research question.

What should not be in an abstract

Do not add data that are absent from the article. Avoid literature citations unless the journal explicitly requires them. Minimise undefined abbreviations because a database reader may see the abstract without the paper's abbreviation list.

Do not claim causation from an observational design. “Was associated with” and “caused” are different claims. Do not hide a negative finding behind “a trend was observed” when the statistical analysis does not support the conclusion.

A weak example

“Remote work is often discussed too generally today. We conducted a study of various aspects of employee productivity. Interesting differences were found between groups. The findings have major practical importance and can be used by companies to improve efficiency.”

This abstract tells us neither the design, sample, measurement nor result. The reader cannot tell what was actually studied.

A rewritten example

“We examined whether fixed team-availability hours were associated with the number of unplanned online meetings in distributed teams. The observational study included 84 teams from four companies; calendar events were analysed over 12 weeks. Teams with a daily two-hour shared availability window held an average of 18% fewer unplanned meetings, while the total duration of scheduled meetings did not differ substantially. The finding shows an association between a predictable contact window and fewer spontaneous calls, but the design cannot establish causality.”

Now the research question, sample, period, primary result and limitation are visible. The final sentence does not advertise the method; it keeps the conclusion within the evidence.

What changed in the revision

“Relevance” was replaced with a concrete research question. “Various aspects” became a measurable outcome. “Interesting differences” became a number. “Practical importance” became a cautious interpretation. The design limitation was added.

ICMJE recommends that an abstract accurately reflect the article and not conflict with it. Many journals use structured abstracts for original research, but the requirements of the chosen journal always take precedence over a generic template.

Write the final abstract after the article is complete

It is useful to rewrite the abstract after the final Results and Discussion are settled. If you drafted it early, the data, terminology and emphasis may have changed. Compare every number with the relevant table and every conclusion with the final Discussion.

Use a reverse test: give the abstract to someone in a neighbouring field and ask whether they can say what was studied, how it was studied and what was found. If they understand only the topic but not the result, the abstract is too general.

Abstract checklist

• the objective fits in one clear sentence;
• design and sample are named;
• the main result is concrete;
• the conclusion is no broader than the data;
• no new facts appear that are absent from the article;
• references are omitted unless the journal requires them;
• abbreviations are minimal and understandable;
• the word count follows the journal limit;
• numbers match tables and Results;
• a limitation is stated when it is needed to prevent overinterpretation.

Structured and unstructured abstracts

If the journal requires headings, do not force every section to have equal length. Methods and Results normally need more space than Background. If the abstract must be one paragraph, preserve the same order of information even without labels.

Systematic reviews, clinical trials and diagnostic studies may have additional reporting requirements. Consult CONSORT, PRISMA, STARD or other relevant standards when they apply to the design.

Allocate the words before you start cutting

Abstracts often become weak not because information is missing but because the author gives equal space to every section of the article. Make a word budget first. In a 250-word abstract, one practical allocation might be 40–50 words for problem and objective, 60–70 for methods, 70–80 for results and 40–50 for the conclusion. This is not a journal rule; it is a way to avoid spending half the abstract on background.

Protect the Results section. If you have to choose between another sentence of context and one concrete result, the number that answers the study objective is usually more valuable. “A significant improvement was observed” tells the reader less than a result with direction and effect size, when the journal permits the detail.

The same framework across disciplines

Clinical research. Problem: evidence about an intervention is incomplete. Method: design, participants and primary outcome. Result: the main endpoint. Conclusion: what the data support and what they do not prove.

Social science. Problem: it is unclear how a group responds to a change. Method: survey, interviews or administrative data. Result: the main association or difference. Conclusion: interpretation within the limits of the design.

Engineering. Problem: an existing method fails under a particular condition. Method: a new design and test procedure. Result: measured improvement over the baseline. Conclusion: the conditions in which the solution is applicable.

Different fields require different details, but the logic remains: the reader should understand the question, how it was answered, the main result and the boundary of the conclusion.

Why “the results are discussed” tells the reader nothing

That phrase only tells us that the paper has a Discussion section. Replace it with a substantive conclusion. Do not write “the data support the hypothesis” if you can say what difference was found. Do not write “the method proved effective” if you have not named the condition and comparator.

Weak ending: “The findings may have practical significance and require further study.”

Stronger: “Participants who used the reminder were more likely to complete the course within eight weeks, but the small sample does not allow us to estimate rare adverse effects.” Now the reader can see both the result and the limitation.

How to shorten an abstract without losing meaning

First remove the history of the field that is not needed to state the research question. Then remove definitions that your intended scientific reader already knows and literature citations that the journal does not require. Next, look for repetition: results are often first described vaguely in words and then repeated with numbers. Keep the more informative version.

Do not shorten by turning prose into telegram language. An abstract should still read as coherent text. “Pts n=84; intervention 6wk; sig diff” saves words by forcing the reader to decode them and may violate journal instructions.

When a structured abstract helps the author

Headings such as Background, Methods, Results and Conclusions force each sentence into the correct role. If a result appears in Background or new information first appears in Conclusions, the structure exposes the problem quickly. Even when the target journal asks for an unstructured abstract, you can draft the four fields first and then remove the headings and connect the sentences.

An example where the conclusion goes beyond the data

Result: at one school, the number of late arrivals fell after a free-breakfast programme was introduced. Too strong: “Free breakfast improves school attendance.” One school and one outcome do not establish that broad causal claim.

More careful: “After the programme was introduced at this school, late arrivals decreased; the study design cannot separate the effect of breakfast from other changes.” A good abstract does not need to make the result louder. Its task is to state exactly what was measured and how strong a conclusion the evidence supports.

If the study has several results

Choose the result that corresponds to the primary objective stated at the beginning. Secondary findings can be added if space remains and they help interpretation. Do not turn the abstract into a catalogue of every table. If the main result does not fit because five secondary outcomes occupy the space, the priorities are wrong.

After writing the abstract, compare every statement with the full paper. No number should appear in the abstract if it is absent from Results, and no causal claim should appear if Discussion avoids it. A common error occurs after late manuscript changes: the table is updated, but the old number remains in the abstract. Final checking should be literal.

Read the abstract separately from the paper a day later. If, without remembering the manuscript, you cannot tell who participated, what was compared and what the primary result was, the abstract relies too heavily on the full-text context. It should be a self-contained short report, not a set of pointers to sections below.

Where to explain the result more simply

An abstract is written for scientific navigation. It saves colleagues time, but it does not have to be good communication for the general public. For a public explanation, write a separate column: begin with the question, remove methodological abbreviations, explain the size of the effect and state the limitation honestly.

What Columnist provides

Science Official lets a researcher publish exactly that kind of popular-science explanation through Columnist. It does not replace the abstract and does not undergo scientific peer review. The author receives a normal publication page, byline and archive of columns, while the reader receives a plain-language explanation of the study.

After preparing the abstract, write a separate explanation for readers and publish it through Columnist, at columnist.mib.news.

FAQ

All guides in the handbook