Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Publication · Editing and proofreading

Editing and Proofreading a Research Paper: A Sequential Four-Pass Approach

Editing a research paper comprises four passes, and performing them in the wrong order wastes most of the work. The one editing intervention that has been evaluated in a randomised trial, review of a manuscript against a reporting checklist, more than doubled the proportion of papers that improved from their submitted state. Sentence-level polishing has no comparable supporting evidence, yet it is the pass that most authors perform first.

Pubrica AcademyWritten by the medical writing and biostatistics teams Reviewed byPubrica biostatistics team Published9 October 2026 Last checked9 October 2026 Reading time10 minutes Sources25 peer-reviewed

Many authors edit by reading the manuscript from the beginning and correcting whatever they notice. This approach produces a paper whose introduction has been rewritten many times while the methods still do not state who held the randomisation sequence. The alternative is a sequence — structure, then science, then sentences, then proof — with separate passes for references and for numbers. The sequence takes longer to describe but less time to perform, and it concentrates effort on the two short texts on which editors most often report basing rejection: the abstract and the title.

Learning objectives

  • The four editing passes, in order, and what belongs in each
  • What the randomised evidence shows about checklist-based review of a manuscript
  • Which sections editors cite as grounds for rejection, and the brevity of those sections
  • Self-editing techniques that are effective where rereading is not
  • What to check in a reference pass that a style check cannot find
  • What may and may not be changed at proof stage

Common misconceptions

  • “Editing means improving the writing.” The intervention supported by randomised evidence is structural: in Cobo and colleagues’ masked trial of 92 manuscripts, review against a reporting checklist increased the proportion of papers improving from baseline from 8 of 41 to 22 of 51 — a difference of 23.6 percentage points and a relative risk of 2.21.
  • “Start at the introduction and work down.” The introduction is the section most likely to be rewritten unnecessarily, because what it must establish depends on what the results support. It should therefore be edited late.
  • “The abstract is a summary you write at the end in ten minutes.” Editors surveyed about desk evaluation reported rejecting on the abstract in 88.09% of cases, and on the title in 28.30%. Per word, these are the most consequential texts in the manuscript.
  • “A reference manager guarantees the references are right.” It guarantees only the formatting. In one audit, major quotation errors — in which the cited source does not support the claim — were found in 39.1% of sampled references.
  • “Proofreading is where you fix the writing.” Proofreading consists of checking the typeset proof against the accepted manuscript. Substantive changes are usually not permitted at that stage, and are sometimes charged for.

Four sequential editing passes

Pass one: structure. The questions at this stage are whether the paper answers a single question; whether the introduction sets up that question and nothing else; whether the methods describe what was done in sufficient detail to be judged rather than trusted; whether the results report the pre-specified outcomes first and in the same order as the methods; and whether the discussion claims only what the results support. The output of this pass is usually a plan to move or cut material rather than a revised sentence.

Pass two: science and reporting. The reporting guideline for the study design — CONSORT for a parallel-group randomised trial, PRISMA 2020 for a systematic review, STARD for diagnostic accuracy, STROBE for observational designs — should be worked through item by item, recording where each item is addressed 8,18. Items that cannot be located represent either a reporting gap that can still be corrected or a conduct gap that cannot, and distinguishing between the two is the most valuable half hour in the entire sequence.

Pass three: sentences and numbers. This is copyediting, and it is most effective as a set of single-purpose sweeps rather than one careful reading: a numbers sweep, a terminology sweep, a statistics-reporting sweep, an abbreviation sweep and a callout sweep. The Academy page on what copyediting catches lists the recurring defects.

Pass four: proof. After acceptance, against the typeset proof, checking for errors introduced in production rather than in the author’s text.

Performing these passes in reverse order is the default and is wasteful. A sentence polished in a paragraph that pass one will delete represents discarded work, and an abstract perfected before the results section is final will not match it.

The four passes: what to do, and what not to do yet

PassWhat you are looking forDo not do this yet
1. StructureOne question; an introduction that sets up that question; methods in the order in which the results are reported; a discussion claiming no more than the results supportSentences in material that may be cut should not yet be polished
2. Science and reportingEvery item of the reporting guideline for the design, located to a section; reporting gaps separated from conduct gapsThe abstract should not be written until this pass is complete
3. Sentences and numbersSingle-purpose sweeps: numbers, terminology, statistical reporting, abbreviations, callouts, referencesNo restructuring; if restructuring is required, return to pass 1
4. ProofWhat production introduced: dropped lines, mis-set symbols, swapped figures, wrong captions, shifted reference numbers, author names and affiliationsNo substantive rewriting; most journals will not permit it

A working sequence, not a research finding. The evidence cited on this page concerns checklist-based review of manuscripts and editors’ reported rejection points.

Evidence from controlled studies

Few of the interventions marketed as editing have been tested. One has: Cobo and colleagues, in a masked randomised trial published in the BMJ, randomised 92 manuscripts submitted to a biomedical journal to receive either an additional review against a reporting checklist or usual review 17. Twenty-two of 51 papers in the intervention arm improved from their baseline state, against 8 of 41 in the control arm — a difference of 23.6 percentage points, a relative risk of 2.21 and a number needed to treat of about four manuscripts to improve one.

The same group had earlier randomised the addition of a statistical reviewer, reported in PLoS ONE, and found measurable improvement in the reporting of accepted articles 20. Day and colleagues’ content analysis in the Annals of Emergency Medicine shows what dedicated methodology and statistics reviewers catch that subject reviewers do not.

Interpreted as practical advice for authors, this is the most useful finding in the editorial literature: what measurably improves a manuscript is systematic review against a structured instrument 17,20. Authors can apply this intervention to their own manuscripts before submission, at no cost, at the checklist stage — and it is a different activity from rereading the paper.

The one randomised trial of checklist-based manuscript review

ArmPapersImproved from baselineEffect
Additional review against a reporting checklist5122—
Usual review (control)418—
Difference92 randomised in total—23.6 percentage points; relative risk 2.21; about 4 manuscripts reviewed to improve 1

Cobo E, Cortes J, Ribera JM, Cardellach F, Selva-O’Callaghan A, Kostov B, et al. Effect of using reporting guidelines during peer review on quality of final manuscripts submitted to a biomedical journal: masked randomised trial. BMJ.

Allocation of editing effort

Saragih and colleagues, in the Malaysian Journal of Library & Information Science, asked journal editors at which points in a manuscript their rejection decisions are formed 3. The abstract was named in 88.09% of cases, the title in 28.30% and the reference list in 27.97% 3. Combined with the finding that editors accept on topic relevance, scope and novelty, the asymmetry is clear: editors accept on properties of the study and reject on properties of the reporting — above all on the shortest piece of reporting in the paper 3,12.

An abstract is 250 words and a title twelve. Together they carry a large share of the editorial decision, yet both are commonly written last, quickly, by whoever has the file open. Where editing time is limited, a disproportionate share of it should be allocated to them: the abstract should state the design, the setting, the population, the primary outcome and an effect size with its uncertainty, and the conclusion should not exceed what that estimate supports 18. Abstract support addresses the structured formats that journals require.

The reference list, at 27.97%, surprises most authors and should be taken literally: an editor who finds the references careless, mis-formatted or implausible for the field interprets this as a signal about the rest of the manuscript.

88.09%of editors named the abstract as a rejection point
28.30%named the title
27.97%named the reference list

Where editors report that rejection decisions form. Source: Saragih E, Saddhono K, Yannuar N, Sumbayak DM. Scholarly Manuscript Assessment: Scientific Journal Editor Desk Evaluation. Malaysian Journal of Library & Information Science.

The same abstract conclusion, edited

Unsupported and unquantified

Conclusions: The nurse-led intervention was effective in improving glycaemic control and should be considered for wider implementation in primary care. Further studies are warranted.

“Effective” with no number. An editor reading only the abstract — which is 88.09% of them, on their own account — cannot see the size of the effect.

A recommendation the design cannot support. A single-centre trial does not support a claim for wider implementation.

No limits. Power, follow-up duration and unmeasured outcomes are all omitted, so the conclusion appears stronger than the study warrants.

“Further studies are warranted”. This statement applies to every published study; it conveys no information and consumes words needed within the word limit.

No setting. Whether a study was single-centre or multicentre is among the first details a reader needs in order to judge generalisability.

Matches what the study can show

Conclusions: In this single-centre trial, the nurse-led intervention reduced HbA1c at 24 weeks by 0.6 percentage points more than usual care (95% CI 0.4 to 0.8). The difference is within the range considered clinically meaningful for this population, but the trial was not powered for microvascular outcomes and follow-up ended at 48 weeks, so durability and clinical benefit remain untested. Replication in multicentre primary-care populations is the next step.

The effect appears with its interval, so size and precision are both visible in the sentence that carries the decision.

The claim is scoped to the design — single-centre, one time point — which makes it credible rather than merely cautious.

The limits that matter are named specifically: not powered for microvascular outcomes, follow-up ended at 48 weeks.

Clinical meaning is addressed separately from statistical difference, which is the distinction reviewers most often find missing.

The next step is specific enough to be actionable rather than formulaic.

The abstract is the text on which editors most often report basing rejection, and it is 250 words long. Rewriting this paragraph carefully is the most valuable hour in the entire editing sequence.

Techniques for self-editing beyond rereading

Rereading one’s own manuscript largely confirms it, because the author supplies from memory what the sentences do not state. Effective techniques interrupt this process by changing what is read or how it is read.

Reverse-outline it. One sentence per paragraph should state the function of that paragraph, and the list should then be read on its own. Paragraphs with two functions, paragraphs with none and sections in the wrong order are evident in the list but invisible in the prose.

Read the tables and figures first, in order, with the captions. If the findings are not evident in them, they are not evident in the paper. Every number in the text should then be checked against the tables for presence and agreement.

Do single-purpose sweeps. Separate passes should be made for numbers, terminology, statistical reporting, abbreviations and callouts. A mixed reading identifies fewer errors of every kind.

Read the methods as a hostile reviewer. For every procedural sentence, the question is what a reader must take on trust. “Patients were randomised” asks for trust; “randomised 1:1 by a computer-generated sequence held by an offsite statistician” does not.

Change the surface. The manuscript may be printed, set in a different font, read aloud or read by text-to-speech software. Reading aloud reveals omitted words and sentences that cannot be completed, both of which the eye corrects silently.

Leave it for 48 hours before the final pass. No other technique restores the necessary distance; a paper set aside for two days reads as though written by someone else.

The reference pass

Two different checks are often confused, and only one of them is time-consuming. The formatting check — style, order, completeness — is performed by reference managers and verified by copyeditors. The accuracy check requires opening the source and confirming that it states what has been claimed, and it cannot be automated.

The time is justified because the error rate is high 16,25. Singh and Chaudhary, auditing sampled references, found major quotation errors — in which the cited source does not support the assertion — in 39.1% of them, and found one particular form of quotation error recurring eleven times in a sample of fifty 19. Such an error is a defect in the evidence rather than in the formatting, and peer review rarely detects it because reviewers seldom retrieve the cited references.

Not every reference needs to be checked. The load-bearing citations should be identified: the two or three in the introduction that establish the gap the study addresses, those supporting methodological choices, and those on which the discussion relies when interpreting the result. These should be checked against the sources, and reference validation used for the formatting pass over the remainder.

Editing in response to peer review

A revision is not a re-edit. It is a response to a set of specific objections, and the editing it requires is driven by the reports rather than by the author’s own view of the paper. Each comment should be sorted into one of three categories — factually answerable, remediable with work that can still be done, or settled when the study was designed — and only what the first two categories require should be edited. The Academy page on rejection and what to do about it sets out that sorting, and response-to-reviewers support is where the response letter is prepared.

Two practical cautions apply. First, every edit made in revision must be reconciled with the rest of the paper: a changed analysis means changed numbers in the abstract, the results, the tables and often the discussion, and a mismatch between old and new numbers is the most common defect in revised manuscripts. Reviewers who have already read the paper are the readers most likely to detect it. Second, the response letter and the manuscript must be consistent: if the letter states that a sensitivity analysis was added, the methods must state this as well.

One finding is relevant when responding to an unfavourable report. Weber and colleagues, in JAMA, found that author satisfaction with peer review depended on the quality of the review as well as on whether the paper was accepted 21. Authors can distinguish a careful critique from a careless one, and the corollary is that the careful critique is the one against which to edit.

Permissible changes at proof stage

Proofreading is a comparison rather than an improvement. The typeset article is checked against the accepted manuscript for errors introduced in production: dropped or doubled lines, mis-set equations and symbols, figures swapped or wrongly oriented, captions attached to the wrong item, corrupted special characters in names, shifted reference numbering, mislinked supplementary files, and incorrectly rendered author names, affiliations and ORCIDs.

Rewriting is generally not permitted. Most journals permit the correction of errors and will refuse or charge for substantive changes, and some will query any alteration that changes meaning. This constraint is the rationale for the sequence described here: sentence-level work must be completed before acceptance, because the proof does not provide an opportunity for it. Proof-stage proofreading and post-acceptance support describe what may reasonably be requested at this stage.

The proof should be checked against the accepted file rather than against memory, tables and equations should be read character by character, and the author list and affiliations should receive a separate reading, since errors there persist in the published record and are the most troublesome to correct afterwards.

Key points

Key points

Editing should proceed in four passes — structure, science and reporting, sentences and numbers, then proof — because work done out of order is work discarded. The only editing-style intervention with randomised evidence behind it is structured: in Cobo and colleagues’ masked trial of 92 manuscripts, review against a reporting checklist raised the share improving from baseline from 8 of 41 to 22 of 51, a relative risk of 2.21, and authors can apply that pass to their own paper before submission. Effort should be concentrated where editors report that decisions are made: the abstract, named by 88.09% of surveyed editors, the title at 28.30% and the reference list at 27.97%. The references on which the argument depends should be checked against the sources, since major quotation errors appeared in 39.1% of one audited sample and no software finds them. Sentence-level work should be completed before acceptance, because the proof is for correcting what production introduced, not for rewriting.

Application to your study

Application to your study

  1. The reporting guideline for the study design should be applied item by item, with the location of each item recorded. This is the pass supported by randomised evidence. Each unmet item should be classified as a reporting gap or a conduct gap.
  2. The paper should be reverse-outlined before any sentence is revised. One line per paragraph should state its function, and the list should be read on its own.
  3. The abstract should be rewritten last, with an hour allocated to it. It should report the design, setting, population, primary outcome, effect size with its interval, and a conclusion that does not extend beyond the data.
  4. Each sweep should address a single type of defect. Numbers, terminology, statistical reporting, abbreviations and callouts should be checked in separate passes, which identify more errors than a single careful reading.
  5. The dozen references on which the argument rests should be checked against the sources. Formatting software cannot determine whether the source supports the claim.
  6. The manuscript should be set aside for 48 hours before the final reading, and its presentation altered. It may be printed, set in a different font, or read aloud.
  7. After peer review, only the changes required by the reports should be made, and every number should then be reconciled again. Mismatches between original and revised figures are the most common defect in revised manuscripts.
  8. At proof stage, the proof should be compared with the accepted file rather than with memory, and the author list should be read separately. Errors at this stage persist in the published record.

Frequently asked questions

Editing takes place before acceptance and changes the manuscript: structure and argument first, then whether the reporting matches the study, then sentences, numbers, terminology and references. Proofreading takes place after acceptance, on the typeset proof, and is a comparison rather than an improvement: it checks for errors introduced in production, such as dropped lines, mis-set symbols, swapped figures, wrong captions, shifted reference numbers and misrendered names. Most journals will refuse or charge for substantive changes at proof stage, which is why all sentence-level work must be completed before submission rather than deferred to the proof.

Structure, then science and reporting, then sentences and numbers, then the proof. The first pass asks whether the paper answers one question, whether the introduction sets up that question, and whether the discussion claims only what the results support; its output is usually a decision to move or cut material. The reporting guideline for the design is then worked through item by item. Only then is the sentence-level work undertaken, as separate single-purpose sweeps. Editing in the reverse order is common, and it results in polishing prose in paragraphs that the structural pass will delete, and perfecting an abstract before the results it summarises are final.

One form of editing has randomised evidence behind it. Cobo and colleagues, in a masked randomised trial in the BMJ, allocated 92 manuscripts to an additional review against a reporting checklist or to usual review: 22 of 51 intervention papers improved from their baseline state, against 8 of 41 controls — a 23.6 percentage point difference, a relative risk of 2.21, and roughly four manuscripts reviewed to improve one. The same group found that adding a statistical reviewer improved reporting in accepted articles. Language polishing as a route to acceptance has no comparable evidence. The defensible claim therefore concerns structured checking against an instrument, not prose.

The abstract, the title and the reference list, in that order, which is counter-intuitive given the brevity of the first two. Saragih and colleagues surveyed journal editors on where rejection decisions are formed: the abstract was named in 88.09% of cases, the title in 28.30% and the reference list in 27.97%. Editors accept on properties of the study — topic relevance, scope, novelty — and reject on properties of the reporting, and the most-read piece of reporting is 250 words long. The methods follow, because that is where a reviewer decides whether to trust the results.

The aim is to interrupt the tendency to read what was intended rather than what was written. The paper should be reverse-outlined: one sentence per paragraph stating its function, with the list then read on its own. Tables and figures should be read first with their captions, and every number in the text checked for presence and agreement in a table. Single-purpose sweeps should replace a single careful reading. The methods should be read with attention to what a reader must take on trust. The presentation should be changed by printing the manuscript, changing the font, reading it aloud or using text-to-speech. Finally, the manuscript should be set aside for 48 hours before the final pass; no other technique restores the necessary distance.

Not every reference, but probably more than most authors check. The formatting check is handled by reference managers and copyeditors. The accuracy check requires opening the source and confirming that it states what was claimed, and the error rate is high: Singh and Chaudhary found major quotation errors, in which the cited source does not support the assertion, in 39.1% of sampled references, with one form of error recurring eleven times in fifty. The load-bearing citations should be checked: those in the introduction that establish the gap, those justifying methodological choices, and those on which the discussion relies. Peer review will not identify these errors, because reviewers rarely retrieve references.

Errors, not writing. The proof pass is a comparison against the accepted manuscript, checking for dropped or doubled lines, mis-set equations and special characters, figures swapped or wrongly oriented, captions attached to the wrong item, shifted reference numbering, mislinked supplementary files, and errors in author names, affiliations and identifiers. Most journals will refuse substantive changes and some will charge for them, and any change that alters meaning will at least be queried. The proof should be read against the accepted file rather than from memory, tables and equations checked character by character, and the author list given a separate pass, since such errors persist in the published record.

One person should be responsible for the file and a different person for the final reading. Comments should be collected in a single version rather than by merging eight annotated copies, and a cut-off should be set after which structural comments are closed. A one-page style sheet — the preferred term for each recurring construct, the spelling convention and the abbreviation list — should be circulated, because inconsistent terminology across sections is the clearest signature of a multi-author draft. Every author must still approve the final version to meet the ICMJE authorship criteria, so time for this should be built into the schedule rather than treated as a formality on the submission form.

References

  1. International Committee of Medical Journal Editors. Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals. International Committee of Medical Journal Editors. 2026. read 9 October 2026. icmje.org
  2. Arsyad S, Zaim M, Othman J. Dealing with journal review results: How successful and productive authors avoid and respond to manuscript rejections. EduLite: Journal of English Education, Literature, and Culture. 2026;11(1):34-48. doi:10.30659/e.11.1.34-48 read 9 October 2026. doi.org/10.30659/e.11.1.34-48
  3. Saragih E, Saddhono K, Yannuar N, Sumbayak DM. Scholarly Manuscript Assessment: Scientific Journal Editor Desk Evaluation. Malaysian Journal of Library & Information Science. 2024;29(2):1-18. doi:10.22452/mjlis.vol29no2.1 read 9 October 2026. doi.org/10.22452/mjlis.vol29no2.1
  4. Yanik B, Evcik D, Geler Kulcu D, Koldas Dogan S, Bardak AN, Zateri C, Sonel Tur B. Why do manuscripts submitted to the Turkish Journal of Physical Medicine and Rehabilitation get rejected? Turkish Journal of Physical Medicine and Rehabilitation. 2023;69(4):535-540. doi:10.5606/tftrd.2023.13204 read 9 October 2026. doi.org/10.5606/tftrd.2023.13204
  5. Reed H, Dehn RW, Bushardt RL. Reasons for unsuccessful research submissions to JAAPA. Journal of the American Academy of Physician Assistants. 2022;35(5):54-56. doi:10.1097/01.JAA.0000824948.31791.0b read 9 October 2026. doi.org/10.1097/01.JAA.0000824948.31791.0b
  6. Menon V, Varadharajan N, Praharaj SK, Ameen S. Why Do Manuscripts Get Rejected? A Content Analysis of Rejection Reports from the Indian Journal of Psychological Medicine. Indian Journal of Psychological Medicine. 2022;44(1):59-65. doi:10.1177/0253717620965845 read 9 October 2026. doi.org/10.1177/0253717620965845
  7. Page MJ, Moher D, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. BMJ. 2021;372:n160. doi:10.1136/bmj.n160 read 9 October 2026. doi.org/10.1136/bmj.n160
  8. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021:n71. doi:10.1136/bmj.n71 read 9 October 2026. doi.org/10.1136/bmj.n71
  9. Lee J, Yang K, Oh DG. Factors influencing the choice of a publication venue in library and information science. Learned Publishing. 2020;33(3):323-332. doi:10.1002/leap.1300 read 9 October 2026. doi.org/10.1002/leap.1300
  10. Editage / UKSG (report summary). Can language editing improve manuscript acceptance rates? UKSG (report summary). 2019. read 9 October 2026. uksg.org
  11. Science Editing (survey report). Authors' perspectives on academic publishing: initial observations from a large-scale global survey. Science Editing. 2018. read 9 October 2026. doaj.org
  12. Meyer HS, Durning SJ, Sklar DP, Maggio LA. Making the first cut: an analysis of academic medicine editors' reasons for not sending manuscripts out for external peer review. Academic Medicine. 2018;93(3):464-470. doi:10.1097/ACM.0000000000001860 read 9 October 2026. doi.org/10.1097/ACM.0000000000001860
  13. Massimo L, Codato A. On the immediate rejection of manuscripts without external peer review. SciELO in Perspective (Revista de Sociologia e Politica). 2016. read 9 October 2026. blog.scielo.org
  14. Walther S, Schueler S, Tackmann R, Schuetz GM, Schlattmann P, Dewey M. Compliance with STARD Checklist among Studies of Coronary CT Angiography: Systematic Review. Radiology. 2014;271(1). doi:10.1148/radiol.13121720 read 9 October 2026. doi.org/10.1148/radiol.13121720
  15. Korevaar DA, van Enst WA, Spijker R, Bossuyt PMM, Hooft L. Reporting quality of diagnostic accuracy studies: a systematic review and meta-analysis of investigations on adherence to STARD. Evidence-Based Medicine. 2014;19(2):47-54. doi:10.1136/eb-2013-101637 read 9 October 2026. doi.org/10.1136/eb-2013-101637
  16. Journal of Guilan University of Medical Sciences (study report). Accuracy of References Used in Research Articles Published By the Journal of Guilan University of Medical Sciences. Journal of Guilan University of Medical Sciences. 2013;22(86):78. read 9 October 2026. magiran.com
  17. Cobo E, Cortes J, Ribera JM, Cardellach F, Selva-O'Callaghan A, Kostov B, et al. Effect of using reporting guidelines during peer review on quality of final manuscripts submitted to a biomedical journal: masked randomised trial. BMJ. 2011;343:d6783. doi:10.1136/bmj.d6783 read 9 October 2026. doi.org/10.1136/bmj.d6783
  18. Schulz KF, Altman DG, Moher D. CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials. Trials. 2010;11(1):32. doi:10.1186/1745-6215-11-32 read 9 October 2026. doi.org/10.1186/1745-6215-11-32
  19. Singh S, Chaudhary R. Accuracy of references cited in articles published in Indian Journal of Dermatology, Venereology and Leprology: A pilot study. Indian Journal of Dermatology, Venereology and Leprology. 2009;75(5):488-491. doi:10.4103/0378-6323.55392 read 9 October 2026. doi.org/10.4103/0378-6323.55392
  20. Cobo E, Selva-O'Callaghan A, Ribera JM, Cardellach F, Dominguez R, Vilardell M. Statistical Reviewers Improve Reporting in Biomedical Articles: A Randomized Trial. PLoS ONE. 2007;2(3):e332. doi:10.1371/journal.pone.0000332 read 9 October 2026. doi.org/10.1371/journal.pone.0000332
  21. Weber EJ, Katz PP, Waeckerle JF, Callahan ML. Author perception of peer review: impact of review quality and acceptance on satisfaction. JAMA. 2002;287(21):2790-2793. read 9 October 2026. garfield.library.upenn.edu
  22. Meadows J. Editorial peer review: its strengths and weaknesses. Journal of Documentation. 2002;58(1):104-106. read 9 October 2026. garfield.library.upenn.edu
  23. Paulus K. How to survive peer review. Learned Publishing. 2002;15(4):312. read 9 October 2026. garfield.library.upenn.edu
  24. DeMaria AN. Peer review: better than the alternatives. Journal of the American College of Cardiology. 2002;40(5):1017-1018. read 9 October 2026. garfield.library.upenn.edu
  25. Ngan Kee WD, Roach VJ, Lau TK. The accuracy of references in the Hong Kong Medical Journal. Hong Kong Medical Journal. 1997;3(4):377-380. read 9 October 2026. hkmj.org

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 9). Editing and Proofreading a Research Paper: A Sequential Four-Pass Approach. Pubrica. https://pubrica.com/academy/editing-service/editing-and-proofreading-your-research-paper/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Four-pass editing schedule

This sheet should be worked through in order, without moving ahead. Each pass has a stop rule: work identified as belonging to a later pass should be noted and deferred. Each pass should be dated so that it can be confirmed that 48 hours elapsed before the final reading.

EDITING SCHEDULE
Manuscript: ______________________  Target journal: ______________________
Reporting guideline for this design: ______________________

PASS 1 - STRUCTURE            date: ____/____/____
  [ ] The paper answers ONE question. Write it here:
      ____________________________________________________
  [ ] The introduction sets up that question and nothing else
  [ ] Methods are in the order the results report them
  [ ] Results report the pre-specified primary outcome FIRST
  [ ] Discussion claims nothing the results do not support
  [ ] Reverse outline written (one line per paragraph)
  Material to cut or move: ___________________________________
  STOP RULE: no sentence polishing in this pass

PASS 2 - SCIENCE AND REPORTING  date: ____/____/____
  [ ] Guideline items worked through one by one
      Items addressed, with section noted: ______ of ______
  Unmet items, each classified:
      ______________________  [ ] reporting gap  [ ] conduct gap
      ______________________  [ ] reporting gap  [ ] conduct gap
  [ ] Registration number stated in methods
  [ ] Analysis population and missing-data handling stated
  [ ] Named test per outcome (no "appropriate tests")
  [ ] Limitations name the specific limits, not generic caution
  STOP RULE: abstract is not written until this pass closes

PASS 3 - SENTENCES AND NUMBERS  date: ____/____/____
  Sweep 1  [ ] Numbers: abstract / text / tables / figures reconcile
  Sweep 2  [ ] Counts with percentages; denominators per group
  Sweep 3  [ ] Estimates with dispersion, intervals, exact p values
  Sweep 4  [ ] Terminology: one name per construct throughout
  Sweep 5  [ ] Abbreviations defined once, then used
  Sweep 6  [ ] Units, conversions, nomenclature conventions
  Sweep 7  [ ] Figure / table callouts match the uploaded files
  Sweep 8  [ ] References: formatting to journal style
  Sweep 9  [ ] References: load-bearing ones checked AGAINST SOURCES
           listed: ______________________________________
  [ ] Abstract rewritten last: design, setting, population,
      primary outcome, effect size with interval, bounded conclusion
  [ ] Title re-read on its own, as an editor would see it first
  [ ] 48 hours left before the final read   gap taken: ______ hours
  [ ] Final read done on paper / in a different font / aloud

PASS 4 - PROOF                  date: ____/____/____
  [ ] Proof compared against the ACCEPTED file, not from memory
  [ ] Equations and special characters checked character by character
  [ ] Figures: correct item, orientation, caption, resolution
  [ ] Reference numbering still aligned with the text
  [ ] Supplementary files linked and open correctly
  [ ] Author names, affiliations, identifiers, corresponding address
  [ ] Only error corrections requested; no rewriting attempted

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica performs the pass supported by randomised evidence: the manuscript is worked through item by item against the reporting guideline for its design, with every unmet item classified as a reporting gap that can still be addressed or a conduct gap that cannot. Pubrica then undertakes the sentence and number work — reconciliation across abstract, text and tables, statistical reporting in a form that can be evaluated, and verification of key references against their sources — and rewrites the abstract last, because it is the text on which editors report basing rejection.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Pubrica returns manuscripts edited pass by pass for the target journal, with the remaining gaps identified.

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.