A Comprehensive Guide to Scientific Writing and Publishing
A paper is an argument with a paper trail. The reader has to find the question, the evidence, and the limit of that evidence without reconstructing them from your notebooks.
Reporting the results is the smaller half of the job. The manuscript also has to show where the work sits in what has already been published, and it has to survive a reviewer who does not share your assumptions. The lists below are the checks that make that possible. They are compiled from writing guides, not from the style rules of one journal.
Words that carry the logic
Connectors are not decoration. They tell the reader whether the next sentence adds, follows, contradicts, or concludes. Pick the relation you mean.
| You are about to | Words that say so |
|---|---|
| Add | Additionally, Furthermore, Moreover, In addition |
| Draw a consequence | Therefore, Thus, As a result, Hence |
| Stress a point | Especially, Certainly, Indeed |
| Give a reason | Because, Due to, Owing to, Since |
| Contrast | However, Although, Whereas, In contrast, Nevertheless |
| Summarize | In summary, To conclude, In short, In brief |
| Give an example | For instance, Such as, Namely, In particular |
“However” and “Therefore” do a lot of work and become invisible when every paragraph starts with one of them. Vary them only when the relation really changes. A contrast word in front of a sentence that does not contrast is worse than repeating “and.”
Discussion and conclusion
The discussion is where the result has to earn its place. A reliable order:
- Answer the question you posed. Restate the finding in the form of that answer.
- Put the answer next to the studies it agrees with and the studies it does not.
- Deal with the other explanation a skeptical reader will reach for.
- Say what the study can and cannot support. Sample, method, and scope belong here.
- Stop on the implication you are willing to defend, or on the next experiment. A forward look is specific: which cohort, which assay, which comparison. “Further studies are needed” is not a conclusion.
Literature review
A literature review that lists papers in the order you read them is a diary. A review that a paper needs is a map.
- Group studies by the claim they share, not by author.
- Say where your question sits on that map.
- Name the gap in terms of evidence, not in terms of your wish to publish.
- Leave the reader with the disagreement that is still open.
The review has a few jobs at once. It shows which results depend on which other results. It puts your question in context. It records which methods have already failed at this problem. It surfaces contradictions instead of averaging them into a smooth paragraph. A table or a concept map of those clusters is often shorter than the prose, and harder to fudge.
Methods someone else can run
The methods section is the part a stranger should be able to follow without emailing you. Include the procedure, the amounts, the times, the settings, the software and dataset versions, the equipment, and the statistical test, including what you treated as the threshold for a claim.
Write it in the past tense, because the experiment has already happened. Name the actor when the actor matters (“we excluded”, “the pipeline dropped”). Passive voice is allowed. It is not a requirement, and it is a bad habit when it hides who made a choice.
Split a long protocol into materials, instrumentation, procedure, and statistics. Move a step-by-step recipe that would swamp the paper into the supplement, and point to it.
Research paper or review
| Research paper | Review | |
|---|---|---|
| Job | Report what you measured or computed | Judge a body of work that already exists |
| Evidence | Primary data | The papers you are reviewing |
| Shape | Introduction, methods, results, discussion | Themes, or time |
| Length often quoted | 3,000–12,000 words | 3,000–5,000 words for a short review |
Those lengths are the ranges this checklist often sees. The journal you submit to publishes its own limit, and a review of a large field is commonly longer than the short-review range.
The sections, one at a time
Abstract
Six sentences are enough for a first draft, in this order. The field. The gap. What you did. The main result, with the number or comparison that makes it a result. What that result means. What it changes. If sentence four has no result in it, the abstract is still a plan.
Results
Walk the figures in the order a reader meets them. Each figure gets one claim in the text. The sentence says what is true in the figure. Interpretation that compares your claim with other papers belongs in the discussion. A results section that is only interpretation has hidden the observations.
Discussion
Lead with the main finding. Then the comparison, the limits, and the move you think the field can make because of this result. The five-part order above is the same thing, written as a checklist.
While you are still drafting
Keep the introduction short enough that the question arrives before the reader has forgotten why they opened the file. Let a figure or a table carry a list you would otherwise write out. Write the results while the analysis is fresh, and write the introduction after you know which result survived. Revise in passes: can a stranger follow the logic, then can anything be cut, then the wording. Someone who did not do the experiment is the right reader for the second pass.
At the end of a writing session, put the paper in one sentence. If that sentence does not name the finding, the draft does not have one yet.
Bottom line
Clear writing is how a reader comes to trust the result. Structure makes the question, the evidence, and the limit findable. The prose is in service of that. A manuscript that is pleasant and vague does not become more publishable by another pass of synonyms.
Credit: adapted from OpenAcademics, Nature, Science Translational Medicine, the MIT Biological Engineering Communication Lab, and other writing guides.
