1. Start with the manuscript as a complete package
Before focusing on sentence-level polish, make sure the manuscript has all of its core pieces. A complete package usually includes a title page, abstract, main text, tables and figures, references, and any supplementary material required by the target outlet. The abstract should match the paper exactly, and every table or figure should be numbered, cited in the text, and understandable in context.
Think of this stage as alignment. If the title says one thing, the abstract implies another, and the results section reports different sample sizes or outcomes, the manuscript is not ready for serious review.
2. Build an abstract that actually represents the paper
Background
Use one or two sentences to establish the specific gap. Avoid generic openings that only say a topic is important.
Methods
Name the design, data or corpus, source count when relevant, and the analytic approach. Readers should know what you actually did.
Findings
State the central pattern or result. Do not replace results with implications or vague statements about significance.
Conclusion
Explain the contribution while keeping the limits of the design visible. Do not make causal claims from non-causal work.
3. Write methods that another researcher can understand
The methods section changes by project type, but the standard is the same: a reader should be able to understand how evidence was selected, produced, analyzed, and interpreted.
| Project type | What the methods should explain |
|---|---|
| Literature review | Databases or sources, search dates and strings, selection criteria, screening process, extraction fields, and synthesis approach. |
| Systematic review / meta-analysis | Protocol, eligibility criteria, information sources, study selection, data collection, effect measures, synthesis plan, and risk-of-bias approach. |
| Quantitative study | Participants or dataset, variables, measurement, preprocessing, missing data, analytic model, software/version, and relevant assumptions. |
| Qualitative study | Sampling or corpus, protocol, coding process, codebook, reflexivity, credibility checks, and how examples were selected. |
| Humanities / theoretical work | Corpus, theoretical lens, unit of interpretation, analytic procedure, and attention to counterexamples or competing readings. |
| Computational project | Data/version, preprocessing, model or tool details, comparison baseline, evaluation metrics, and reproducibility materials. |
4. Let the results answer the question
Organize results by the research question or analytic claim, not simply by the order in which data were collected. Report denominators and units when they matter. Figures and tables should have captions that tell the reader what they are seeing, and null or mixed findings should be reported as honestly as positive findings.
5. Use the discussion to interpret, not exaggerate
Start the discussion by answering the research question directly. Then explain what the findings mean, how they fit with prior work, and where the study is limited. Strong limitations sections name the constraints created by design, measurement, sample or corpus, analysis, and generalizability rather than offering a generic sentence about “future research.”
Match the strength of the claim to the strength of the evidence
| If the evidence shows... | Safer wording | Avoid |
|---|---|---|
| A cross-sectional or observational pattern | “is associated with” | Causal language such as “causes” or “improves.” |
| A theory with suggestive evidence | “may contribute to” | Definitive claims. |
| An experiment or quasi-experiment with a real comparison | “improves” may be appropriate if the design supports it | Treating a simple pre/post change as proof of causation. |
| Survey or interview reports | “students reported...” | Describing self-report as objective behavior. |
| A pattern across multiple sources | “the literature suggests...” | Using one isolated citation as though it represents an entire field. |
6. Prepare the manuscript for external submission
Before the paper leaves your team, verify that the references match the in-text citations, terminology is consistent across the manuscript and figures, and the abstract reports the same counts and values as the results. Then check the target journal or conference for word limits, figure specifications, disclosure requirements, author instructions, and supplementary-file rules. A strong paper can still be delayed or rejected if the submission package does not meet the outlet’s requirements.
Helpful standards & references
These are starting points, not substitutes for the requirements of a specific journal, conference, institution, or research team.
ICMJE Recommendations: https://www.icmje.org/recommendations/
EQUATOR Network reporting guidelines: https://www.equator-network.org/reporting-guidelines/
ICMJE Preparing for Submission: https://www.icmje.org/recommendations/browse/manuscript-preparation/preparing-for-submission.html