When we edit a research paper, there are usually plenty of sentences that can be improved. Some are grammatically incorrect. Others are awkward because they have been translated too literally from another language. There may also be sentences that are grammatically correct but difficult to understand.
Most of these problems can be resolved once we understand what the author is trying to say.
The more difficult cases are those in which improving the English can also change the scientific meaning.
Consider the following sentence:
Higher salt intake was associated with increased blood pressure.
An editor may be tempted to make this more direct:
Higher salt intake increased blood pressure.
The revised sentence is shorter and reads more naturally. However, it may no longer accurately describe the study.
“Was associated with” tells us that a relationship was observed between salt intake and blood pressure. “Increased” suggests that higher salt intake caused the increase in blood pressure. Whether the researchers can make that causal claim depends on the study design and what the evidence actually shows.
A small change in wording can therefore change the conclusion that a reader takes from the paper.
This is one of the reasons why editing research involves more than correcting the English.
When a clearer sentence becomes a stronger claim
Research papers often contain cautious language. Authors write that their findings suggest something, that an intervention may have an effect, or that one variable was associated with another.
Sometimes this caution is unnecessary. Academic writing can become overly tentative, particularly when several qualifying expressions are used together.
For example:
It may perhaps be suggested that the intervention could have contributed to improved adherence.
Depending on the findings, this may be unnecessarily cautious and could probably be expressed more clearly.
However, we should not assume that every cautious expression is simply a language problem.
Consider:
Our findings suggest that the intervention may improve treatment adherence.
This could be revised to:
Our findings show that the intervention improves treatment adherence.
The second sentence makes a considerably stronger claim.
Before making such a change, we need to know whether the study actually supports it. What type of study was conducted? Was adherence measured as an outcome? How certain are the findings? Are there limitations that explain why the authors have chosen more cautious wording?
These are not questions that can be answered from the grammar alone.
This does not mean that an editor should make scientific decisions for the author. It means that we need to recognize when a proposed language change could affect the interpretation of the research.
Statistical language can also be changed too easily
We often see the same problem with statements about statistical significance.
An author may write:
There was no statistically significant difference in mortality between the two groups.
It may seem reasonable to simplify this to:
There was no difference in mortality between the two groups.
However, the two statements do not necessarily mean the same thing.
A failure to find a statistically significant difference does not establish that there was no difference at all. The effect estimate, confidence interval, sample size, study design, and other aspects of the analysis can all affect how the finding should be interpreted.
The American Statistical Association has cautioned against using a p value threshold as the sole basis for a scientific conclusion. Statistical significance also does not tell us the size or importance of an effect.
For an editor, this is important because phrases such as statistically significant may sometimes appear repetitive or cumbersome. Removing them simply to make the sentence flow better can produce a statement that is broader than the analysis supports.
Technical terms should not be replaced simply because another word sounds more natural
A similar problem arises when a technical term seems unnecessarily complicated.
For example:
The odds of postoperative complications were higher in Group A.
An editor may feel that risk sounds more natural:
The risk of postoperative complications was higher in Group A.
Whether this change is appropriate depends on what the researchers actually calculated. Odds and risk are related concepts, but they are not the same statistical measure.
There are many other pairs of terms that can appear similar in ordinary language but have different meanings in research. These include:
- incidence and prevalence
- correlation and agreement
- mortality and survival
- relative risk and absolute risk
- predictor and cause
There will also be occasions when an author has used the wrong technical term and it should be corrected. The important point is that we need to establish this from the study rather than replace the term simply because another word sounds better.
Sometimes an edit makes the finding broader than it was
Another issue arises when the revised wording moves away from what was actually measured.
Suppose a study assessed participants using a questionnaire on self-reported sleep quality. The manuscript states:
Self-reported sleep quality improved after the intervention.
This could be shortened to:
Sleep improved after the intervention.
For some types of writing, that might be a reasonable simplification. In a research paper, the distinction matters.
The researchers measured a particular aspect of sleep using a particular method. They did not necessarily establish that sleep improved in every respect.
Similar issues arise when:
- depressive symptom scores become depression
- self-reported physical activity becomes physical activity
- markers of inflammation become inflammation
- a result observed in one study population is written as though it applies to patients generally.
Some of the words that appear unnecessary in the original sentence may actually define the limits of the finding.
Reporting guidelines such as STROBE also ask authors of observational studies to interpret their results in light of the study objectives, limitations, other analyses, relevant evidence, and the generalizability of the findings.
What should we do when the English is wrong but the meaning is not clear?
This is one of the more difficult situations in research editing.
Consider:
Hypertension was an influence factor for postoperative complications.
The sentence clearly needs to be revised. The problem is that several interpretations are possible.
The authors may mean that hypertension:
- was independently associated with postoperative complications in a multivariable analysis
- showed an association in an univariable analysis
- was associated with an increased risk of complications
- was more common among patients who developed complications.
These findings would not all be written in the same way.
If the rest of the manuscript makes the intended meaning clear, the editor can revise the sentence accordingly.
If it does not, choosing what seems to be the most likely interpretation can introduce a finding that the authors did not intend to report.
In that situation, we would normally ask the author to clarify what they mean.
Authors may sometimes wonder why an editor has left a comment instead of simply correcting the sentence. However, if the underlying scientific meaning is uncertain, providing fluent English does not resolve the problem. It can instead make an incorrect interpretation look convincing.
AI makes this issue more important
Generative AI can now turn awkward academic English into fluent prose very quickly. For researchers who work primarily in another language, this can be extremely useful.
The difficulty is that fluent language can make it harder to notice that the meaning has changed.
An AI system may infer what a sentence probably means and rewrite it on that basis. In many situations, that is useful. In research writing, however, there are cases where we should not make that inference.
If a sentence could reasonably have two different scientific meanings, we need information from the paper or from the author before deciding which interpretation is correct.
The International Committee of Medical Journal Editors recognizes that AI tools may be useful to people writing in a language that is not their primary language. At the same time, its recommendations make clear that people remain responsible for checking the accuracy and validity of AI-assisted content.
This does not mean that AI should not be used for language support. It means that improving the language and determining what the research means are not always the same task.
What should research editing achieve?
Most language changes in a research paper will not create any of these problems. If the author’s intended meaning is clear, substantial improvements can be made to the grammar, structure, readability, and overall flow of the manuscript without affecting the research.
The difficult cases are those where the editor cannot be certain that the revised sentence still represents the same finding.
We therefore need to pay attention to more than whether a sentence sounds natural. A revision may change the level of certainty, introduce causality, broaden an outcome, change a statistical measure, or extend a finding beyond the population that was studied.
When that happens, the question is no longer simply how the sentence can be improved. We first need to be sure that we understand what the author intended to say.
If we do not, asking the author for clarification may be the most appropriate option.
Research authors should expect an editor to improve their English. They should also be able to expect that the edited manuscript continues to represent the research they conducted and the findings they obtained.
The aim is to make the research clearer to the reader without changing what the evidence actually shows.
Sources and further reading
- International Committee of Medical Journal Editors (ICMJE). Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals. Updated January 2026.
- STROBE Initiative. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement.
- American Statistical Association. Statement on Statistical Significance and P-Values.
The manuscript examples in this article were created for explanatory purposes. They are not taken from client manuscripts or unpublished research.