Separate a correction from a scientific decision
Start by labelling the problem. Grammar and punctuation can often be corrected directly. An unclear population, measurement or claim needs an author query. Keep a list of those questions beside the edited draft, with the manuscript location and the information needed to resolve each one. An unanswered query is preferable to a fluent sentence containing an invented detail.
Improve the sentence without inventing a method
In the first example, the edit removes awkward grammar and makes the follow-up phrase easier to read. It does not add a follow-up interval, measurement tool or clinical outcome. If the surrounding manuscript does not explain those details, ask the author separately. Do not assume that a phrase such as regular follow-up means a specific visit schedule.
Make uncertainty visible to the author
The second example cannot be completed responsibly from language alone. The word normal has several possible meanings, and a reader needs to know which one applies. A useful query names the missing information and points to its location. The author can then check the protocol, instrument manual or original records and supply a supported answer.
Check the strength of the conclusion
In the third example, the sample sentence claims causation although the hypothetical design is cross-sectional. The alternative describes an association. The author must still confirm that it reflects the actual analysis. The ICMJE recommends conclusions that stay within the evidence presented; polishing language does not remove a study limitation.
Review the edited file as an author
Read every changed sentence in context. Check that numbers, units, group names and directions of comparison still match your tables. Resolve comments before preparing a clean version, then compare the final abstract with the main text. Keep the tracked version so you can explain substantive changes to coauthors. This is a final consistency pass, not a new analysis.
Record the assistance accurately
Agree on the permitted support before sending a draft and check the rules of your institution and journal. ICMJE distinguishes authorship from other contributions: writing or language assistance alone does not automatically qualify someone as an author. Describe assistance as required and obtain any necessary permission to name contributors. The authors remain responsible for the manuscript they submit.
Worked examples
These invented teaching examples show an editing decision. They are not client records or research results.
- Before
- The patients which received treatment were assessed at follow up.
- After or author query
- Patients who received treatment were assessed at follow-up.
A language edit preserves the supplied meaning. No timing or outcome was added.
- Before
- The values were normal in most cases.
- After or author query
- Author query: Which measurement, unit and reference interval does “normal” refer to, and how many observations were assessed?
A query requests the missing facts without filling them in.
- Before
- Our cross-sectional survey proves that shorter sleep causes distress.
- After or author query
- In this cross-sectional survey, shorter sleep was associated with distress.
This hypothetical revision removes a causal claim; the author must confirm the association and its uncertainty from the actual analysis.