Pre-submission manuscript review — 12 things reviewers actually look for

Peer review is a structured form of scepticism. Reviewers are reading your manuscript looking for reasons to recommend rejection — not because they are hostile, but because that is the job. Most of those reasons cluster into a small number of categories, and most of them are visible to you before you submit. This checklist works through the twelve that account for the majority of avoidable rejections.

It applies most directly to empirical research articles in the social, health, and life sciences, but the principles transfer to almost any structured academic writing. Run through it in order. Each item takes a few minutes. The whole pass should take a couple of hours for a typical paper.

01Title

The title is your most-read sentence. It does two things at once: it tells a human reader what the paper is about, and it tells a search engine which queries should surface it. Both matter — discoverability and accuracy.

Check that the title contains the population, the variable of interest, and the design or outcome. "Effect of high-intensity interval training on VO₂ max in adolescent female footballers: a randomised crossover trial" is doing all of that work in one sentence. "A novel approach to fitness training" is doing none of it.

Common error Over-claiming. If the title says "improves performance" but the results show a small effect on one secondary outcome, reviewers will notice and the editor will be irritated. Match the title to what the data actually show.

02Abstract

The abstract is the second-most-read part of the paper, and the only part that most readers will ever see. Most journals enforce a 250-word limit and a structured format: background, methods, results, conclusions. Fit your work into that structure even if the journal does not require it.

The most common abstract problem is what I think of as the methodological black hole: the abstract says "we conducted a study" without specifying the design, sample size, or analytical approach. A reviewer skimming the abstract should be able to answer: who, how many, what design, what outcome, what test, what direction of effect, and how large.

If a number appears anywhere in the abstract, it should match a number that appears in the results section. Mismatches between abstract figures and full-text figures are flagged routinely by attentive reviewers and are one of the most embarrassing things to be asked about in revisions.

03Introduction

The introduction has one job: by the time the reader reaches the aim, they should feel that this study is the obvious thing to do next. That is a higher bar than it sounds.

The conventional structure is a funnel: broad context, narrower context, what is known, what is not yet known, and the aim that addresses the gap. Most weak introductions fail at the second-to-last step. They describe what other people have done at length, then jump to "the aim of this study was..." without making the case that the gap matters or that this is the right way to fill it.

Test the introduction by reading only the last paragraph. If a reader who has not read the rest of the introduction would not see the aim coming, the funnel is not closing properly.

Useful test Ask a colleague outside your subfield to read only the introduction and predict what the methods section will look like. If their prediction is roughly right, the introduction is doing its job.

04Methods reproducibility

The methods section is where reviewers most commonly raise concerns serious enough to recommend rejection. The standard a reviewer applies is: could a competent researcher in this field reproduce this study using only what is written here? If the answer is no, the section is incomplete.

Specific things to check:

• Equipment — manufacturer, model, country. Software needs version numbers. SPSS is not enough; SPSS v29 (IBM, USA) is.
• Sample size justification — either an a priori power calculation with stated effect size, alpha, and power, or a clear statement that the sample was a convenience sample and what the resulting power implications are.
• Procedures — written in sufficient detail that the next research group could replicate them. A reader should not have to chase three older papers to work out what was actually done.
• Data handling — what was excluded, why, and at what stage. Pre-registration of exclusion criteria, where applicable, strengthens this considerably.

05Statistical reporting

Three things should appear for every inferential test: the test name, the effect size, and the confidence interval. P-values alone are no longer sufficient in most fields, and have not been for some years.

Check that the statistical test used is appropriate for the data type and the design. Repeated measures need a repeated-measures analysis. Multiple comparisons need correction. Non-normal data either need a non-parametric test or a robust justification for why parametric methods are still appropriate. Assumption checks should be reported, not just performed.

Report exact p-values to three decimal places where above 0.001, and as p < 0.001 below that threshold. Avoid p = 0.000, which is a reporting convention that no longer survives review.

Common error Reporting only p-values for a primary outcome while burying effect sizes in supplementary material. Reviewers and editors increasingly want effect sizes in the main text, and journals are tightening these requirements.

06Results and discussion separation

The results section reports what you found. The discussion explains what it means. Mixing the two is one of the most common structural issues in submitted manuscripts.

In the results, you state the finding with the supporting statistics. You do not interpret it, you do not compare it to prior literature, and you do not explain mechanisms. All of that belongs in the discussion. A reader of a results section should know what happened in the study without yet knowing what to think about it.

Conversely, a discussion section that repeats results without interpretation is wasting space. Each paragraph in the discussion should add a layer the results section did not — mechanism, comparison with prior work, methodological caveat, or implication.

07Limitations

Every study has limitations. A paragraph that claims there are none is a red flag to a reviewer. So is a paragraph that lists limitations so severe that the paper could not have been worth doing — that reads as performative.

The right tone is honest specificity. Name the limitation, explain what its effect on the findings might be, and note what would mitigate it in future work. Three to five limitations of moderate severity, each handled in a sentence or two, is the typical shape of a well-balanced paragraph.

If a limitation is fundamental — a small sample, a design that cannot establish causality, a measurement instrument with known reliability concerns — name it directly. Reviewers respect candour and will often score the manuscript more leniently when authors are transparent than when they try to obscure known weaknesses.

08Novelty framing

Reviewers and editors will look for an explicit answer to the question what does this paper add that we did not already know? If the answer is buried, vague, or absent, the paper risks being rejected for insufficient contribution even when the methodology is sound.

The novelty claim should be specific and defensible. "This is the first study to..." is a high-risk phrasing — it invites the reviewer to find a prior study you missed, and they often will. Safer phrasings frame the contribution: "adds to a small literature by examining...", "extends prior findings by addressing...", "provides the first evidence in this population for...".

Place the novelty statement at two points: late in the introduction (to set up why the study was worth doing), and early in the discussion (to remind the reader what was contributed). These two statements should be consistent with each other.

09Figures and tables

Every figure and table should be able to stand alone. A reader should be able to look at the figure, read the caption, and understand what is being shown without going back to the main text. If they cannot, the caption needs more information.

• Captions — describe what is shown, what each element represents, sample size, statistical test where relevant, and what the asterisks or shading mean.
• Axes — labelled with both the variable name and the unit. A y-axis labelled "Force" is incomplete; "Vertical ground reaction force (N)" is complete.
• Resolution — at least 300 dpi for raster images, vector format where possible. Pixelated figures suggest a rushed submission.
• Colour — colour-blind safe palettes where possible, and never relying on colour alone to distinguish categories (use shape or pattern as well).

Count your figures. Most journals limit you to five or six. If you have eight, the reviewer is going to ask why, and the editor may ask you to consolidate.

10References

Reviewers do read reference lists, particularly to check whether the work is properly situated in the field. Specific things they look for:

• Recency — for most fields, the majority of citations should be from the last ten years, with a strong cluster from the last five. A reference list dominated by twenty-year-old papers signals that the literature review may be out of date.
• Self-citation balance — citing your own prior work where genuinely relevant is fine. Citing it heavily in a way that excludes other authors in the same area is flagged routinely.
• Reference accuracy — every reference should be findable. Check DOIs work. AI tools are particularly prone to inventing plausible-looking citations that do not exist, so if you have used any AI assistance during writing, every reference needs human verification.
• Citation alignment — what the cited paper actually says should match what your text claims it says. Reviewers spot-check this.

Critical Hallucinated citations from AI tools are now the single fastest route to losing reviewer trust. Verify every reference manually.

11Ethical statements

Most journals require explicit statements covering ethical approval, informed consent, conflicts of interest, funding sources, data availability, and contributor roles. Missing or vague statements in any of these areas can delay review or trigger desk rejection.

For ethical approval, name the committee, the institution, and the reference number. For consent, state the form of consent obtained. For data availability, follow the FAIR principles where possible — even if data cannot be shared openly, state where they are held and the conditions under which they could be made available.

Conflict of interest declarations should be specific. "The authors declare no conflicts of interest" is fine when true. When there are conflicts, name them directly. Vague or apparently incomplete declarations are increasingly flagged at editorial screening stage.

12Cover letter and journal fit

The cover letter is the editor's first read. It should do three things in under 300 words: state what the paper reports, explain why it is a fit for this specific journal (not just journals in general), and confirm any disclosures required by the submission system.

Journal fit is the most common reason for desk rejection — more common than methodological issues. Editors at general-audience journals reject papers that read as too narrow or too technical; specialist journals reject papers that read as off-topic. Check the journal's recent issues, find two or three papers in adjacent areas, and frame your contribution in language that fits how those papers are positioned.

Avoid generic phrasing. "We believe this paper will be of interest to your readers" is empty. "This paper extends [Author 2024]'s findings on X by addressing Y in a clinical population, and we think it sits well alongside your recent special issue on Z" demonstrates that you have read the journal and know where your work fits.