Table of Contents

What Is Scientific Writing?

Q: How long should a scientific paper be?

Length varies enormously by journal and field. A brief communication or letter might be 1,500-2,500 words. A standard research article typically runs 3,000-6,000 words. Review articles can exceed 10,000 words. Most journals specify word limits in their author guidelines. The real answer is: as long as it needs to be to communicate the findings clearly, and not a word longer. Brevity is consistently valued.

Q: What's the difference between a research article and a review article?

A research article (also called an original article or primary literature) reports new findings from original experiments or studies. A review article synthesizes and summarizes existing research on a topic, often identifying patterns, gaps, and future directions. Review articles don't present new data but add value by organizing and interpreting the existing body of knowledge. Both undergo peer review, but they serve different purposes.

Q: Who should be listed as an author on a scientific paper?

According to ICMJE criteria (the most widely adopted standard), authorship requires all four of: (1) substantial contribution to conception/design or data acquisition/analysis, (2) drafting or critically revising the manuscript, (3) approving the final version, and (4) agreeing to be accountable for all aspects of the work. Simply providing funding, lab space, or data collection assistance doesn't qualify for authorship—those contributions belong in the acknowledgments section.

Q: Is passive voice required in scientific writing?

No, and this is a common misconception. While passive voice was traditionally dominant ('the solution was heated' rather than 'we heated the solution'), most major journals and style guides now accept and even prefer active voice. Active voice is clearer, more direct, and easier to read. Use passive voice when the actor is unknown or irrelevant, and active voice when clarity benefits from identifying who did what.

Q: What is peer review and how long does it take?

Peer review is the evaluation of a submitted manuscript by independent experts in the field. Reviewers assess the methodology, analysis, interpretation, and significance of the work. The process typically takes 2-6 months from submission to first decision, though some journals offer expedited review. Reviewers may recommend acceptance, minor revisions, major revisions, or rejection. Most papers that are eventually published go through at least one round of revisions.

Scientific writing is a form of technical communication that reports the methods, results, and interpretation of scientific research to other scientists and, increasingly, to the broader public. Its defining characteristics are precision, clarity, logical structure, and evidence-based argumentation. Every sentence exists to convey specific information, and every claim is either supported by data or properly attributed to its source.

About 3 million scientific papers are published annually across roughly 30,000 peer-reviewed journals worldwide. That number has been growing at approximately 5-6% per year for decades. The volume creates both opportunity and challenge: more knowledge is being shared than ever before, but finding and reading the relevant fraction of it is a real problem. Writing clearly and concisely isn’t just a nicety—it’s an ethical obligation to the readers who are investing their limited time in your work.

Why Scientific Writing Has Its Own Rules

Scientific writing isn’t creative writing with lab data inserted. It’s a distinct genre with conventions that exist for specific reasons. Understanding those reasons—rather than blindly following the rules—is the difference between writing that merely complies and writing that actually communicates.

Precision Over Elegance

In a novel, you might describe rain as “the sky weeping.” In scientific writing, you’d write “precipitation of 23.4 mm over 6 hours.” The novel version evokes feeling; the scientific version enables replication. Another researcher in another country needs to know exactly what happened so they can reproduce your conditions or compare their results to yours.

This precision extends to word choice. “Significant” in everyday English means “important.” In scientific writing, “significant” has a specific statistical meaning (p < 0.05, typically). Using it casually causes confusion. Similarly, “prove” is almost never appropriate in science (you provide evidence for or against; you don’t prove). “Correlation” doesn’t mean “causation.” These distinctions matter because imprecise language leads to imprecise thinking and misinterpretation of findings.

Reproducibility as a Core Value

The scientific method depends on other researchers being able to repeat your work. Scientific writing exists partly to make this possible. Your Methods section should be detailed enough that a competent researcher in your field could reproduce your experiment without contacting you. If they can’t—because you omitted a key step, used ambiguous descriptions, or failed to specify your materials—your paper has failed one of its primary functions.

Attribution and Building on Prior Work

Science is cumulative. Every new finding builds on previous work. Citations serve the dual purpose of giving credit and letting readers trace the intellectual lineage of ideas. When you cite a paper, you’re saying “this claim isn’t mine—here’s where it came from, and you can verify it.” Failing to cite properly is plagiarism. Citing inaccurately—misrepresenting what a source actually says—is arguably worse.

The IMRAD Structure

Most research articles follow the IMRAD format: Introduction, Methods, Results, and Discussion. This structure has been the standard since the 1970s, and journals overwhelmingly require it. It works because it mirrors how science actually proceeds and because readers know exactly where to find specific information.

Introduction: Why This Matters

The Introduction does three things in sequence:

Establishes context: What’s the general topic? Why is it important? What do we already know?
Identifies the gap: What don’t we know? What question remains unanswered? What problem hasn’t been solved?
States the purpose: What does this paper do to fill that gap?

This funnel structure—broad to narrow—is nearly universal. It guides the reader from general knowledge to the specific question your paper addresses. A well-written Introduction makes the reader think “yes, that’s an important question and I want to know the answer”—which is exactly the motivation needed to keep reading.

Common mistakes in Introductions: being too broad (“Since the dawn of time, humans have wondered about…”), being too narrow (jumping to specific details without context), reviewing too much literature (the Introduction isn’t a review article), and burying the purpose statement.

The best Introductions are 2-4 paragraphs for a standard paper. They cite 15-30 references. They end with a clear statement of what the paper does and, sometimes, a brief preview of the key findings.

Methods: What You Did

The Methods section describes your experimental procedures, materials, subjects, equipment, and analytical approaches in enough detail for replication. It’s written in past tense (you’re describing what you did) and traditionally used passive voice (“samples were collected”), though active voice (“we collected samples”) is now widely accepted.

Subsections organize the Methods logically: Study design, Participants/Subjects, Materials, Procedures, Data analysis. Each subsection should be self-contained.

Specificity is non-negotiable. Not “the solution was heated” but “the solution was heated to 95 plus or minus 2 degrees Celsius in a water bath (Model XYZ, Manufacturer) for 30 minutes.” Not “antibodies were used” but “anti-GFP antibody (clone JL-8, Takara, catalog #632381) at 1:1000 dilution.”

Statistical methods get their own subsection. Specify which tests you used, why you chose them, what software performed the analysis, and what significance threshold you applied. Reviewers and readers will scrutinize your statistical choices—as they should.

For computational work, including algorithms used in data processing, provide enough detail (or references to published methods) for replication. Ideally, share your code in a public repository.

Results: What You Found

The Results section presents your findings without interpreting them. That’s the rule, though in practice, some degree of interpretation creeps in, and many journals tolerate or even encourage combined Results and Discussion sections.

Organize by experiment or question, not chronologically. If you ran experiments A, B, and C, but B’s results logically precede A’s in telling your story, present them in the order B, A, C. The reader’s understanding matters more than the calendar.

Text and figures work together. The text highlights the most important findings and guides the reader through the data. Figures and tables present the data itself. Never simply repeat in text what a figure already shows—add value by pointing out key comparisons, trends, or unexpected results.

Report negative results. If an experiment didn’t work or showed no significant effect, say so. Omitting negative results creates publication bias and misleads other researchers. The sentence “Treatment X did not significantly affect outcome Y (p = 0.43, n = 50)” is more valuable than silence because it saves other researchers from pursuing a dead end.

Statistical reporting follows field-specific conventions. In biomedical writing: “Mean blood pressure decreased by 8.3 mmHg (95% CI: 5.1-11.5, p = 0.002).” In psychology: “Participants in the experimental condition scored higher (M = 72.3, SD = 8.1) than controls (M = 65.7, SD = 7.9), t(98) = 4.13, p < .001, d = 0.83.” Data analysis conventions vary between fields, so check your target journal’s requirements.

Discussion: What It Means

The Discussion interprets your results. It’s where you argue for the significance of your findings, compare them to previous work, acknowledge limitations, and suggest future directions.

A strong Discussion follows an inverted funnel—narrow to broad, the mirror image of the Introduction:

Summarize key findings (briefly—don’t repeat the Results section)
Interpret and compare: What do your results mean? How do they relate to previous studies? Do they support or contradict existing theories?
Address limitations: What are the weaknesses of your study? What can’t your results tell us? Being honest about limitations increases credibility—reviewers will identify them anyway
Broader implications: What does this mean for the field? What should future research investigate?

Common mistakes: overstating your results (“this definitively proves”), ignoring contradictory evidence, failing to acknowledge limitations, and speculating far beyond what your data support. The Discussion is where cognitive bias most easily infiltrates scientific writing—the temptation to present your results in the best possible light is real and must be actively resisted.

Other Sections

Abstract: A standalone summary (150-300 words) covering purpose, methods, key results, and conclusion. Many readers will only ever read your abstract, so it needs to stand on its own. Structured abstracts (with labeled subsections) are increasingly required in biomedical journals.

Title: Your paper’s first impression. A good title is specific, informative, and concise. “Effect of Temperature on Enzyme X Activity in Y Organism” beats “A Study of Enzymes” (too vague) and “Investigating the Complex Relationship Between Ambient Environmental Temperature Fluctuations and the Catalytic Efficiency of Recombinant Enzyme X…” (too long).

Keywords: 4-8 terms that help databases index your paper. Choose terms your target audience would search for.

Acknowledgments: Thank people who contributed but don’t meet authorship criteria, and identify funding sources.

References: The list of works you cited. Format varies by journal (APA, Vancouver, Chicago, etc.) but the information is always the same: who wrote it, when, what it’s called, and where to find it.

Writing Style: The Principles

Clarity Above All

If a reader has to re-read your sentence to understand it, you’ve failed. Scientific content is inherently complex; your writing shouldn’t add additional complexity. Short sentences are generally clearer than long ones. Simple words are better than fancy synonyms. Active constructions are usually clearer than passive ones.

Bad: “The determination of protein concentration was carried out utilizing the Bradford assay methodology.” Better: “We measured protein concentration using the Bradford assay.”

The second version is 40% shorter, more direct, and clearer. No information is lost.

Conciseness

Every word should earn its place. Scientific writing suffers from bloat—hedging phrases (“it is interesting to note that”), redundancies (“a total of 50 subjects”), and filler words (“basically,” “actually,” “very”) that add length without meaning.

Academic writing culture sometimes rewards wordiness, creating a perverse incentive to pad papers. Resist this. Reviewers and editors consistently value concise writing. A paper that makes its point in 3,000 words is better than one that takes 5,000 to say the same thing.

Logical Flow

Each sentence should follow logically from the previous one. Each paragraph should have a clear topic sentence and develop a single idea. Transitions between paragraphs should make the logical connection explicit.

Scientific papers often use signposting language: “To test this hypothesis, we…”, “These results suggest…”, “In contrast to previous findings…”, “Taken together, these data indicate…”. This language is sometimes criticized as formulaic, but it works because it tells the reader exactly where they are in the argument.

Tense Conventions

Past tense for describing what was done and what was found (Methods and Results): “We measured…”, “The results showed…”
Present tense for established knowledge and general statements: “Enzyme X catalyzes…”, “Previous studies show…”
Future or conditional tense for implications and proposed future work: “These findings could inform…”, “Future studies should investigate…”

Figures and Tables

Good figures can communicate results more effectively than any amount of text. Bad figures waste space and confuse readers.

Figure Design Principles

One message per figure. A figure should answer one question or make one comparison. If you’re cramming multiple unrelated datasets into a single panel, split them.

Labels and legends must be self-explanatory. A reader should be able to understand your figure without reading the main text. Include axis labels with units, define all symbols in the legend, and add scale bars where relevant.

Choose the right plot type. Bar charts for categorical comparisons. Scatter plots for continuous variable relationships. Line graphs for time series or dose-response curves. Box plots for distributions. Don’t use pie charts for scientific data—they’re notoriously hard to read accurately.

Show the data. Wherever possible, show individual data points rather than just summary statistics (means and error bars). A bar chart with error bars can hide bimodal distributions, outliers, and other patterns that scatter plot overlays reveal.

Color carefully. About 8% of men have some form of color vision deficiency. Use colorblind-friendly palettes (avoiding red-green combinations), and ensure your figures are interpretable in grayscale. Journals still print some content in black and white.

Tables

Tables work best for precise numerical comparisons when exact values matter more than visual patterns. Use them for demographic data, experimental conditions, statistical test results, and quantitative comparisons. Keep tables simple—if a table has more than 5-6 columns, consider splitting it or converting some data to figures.

The Publishing Process

Choosing a Journal

Before writing, identify your target journal. Each journal has specific scope, audience, format requirements, and impact level. Submitting a molecular biology paper to a physics journal wastes everyone’s time. Consider:

Scope: Does the journal publish this type of research?
Audience: Will the right people read it here?
Impact factor: How widely read and cited is the journal? (But don’t obsess over this metric—it has well-documented problems)
Open access: Will your paper be freely available, or behind a paywall?
Turnaround time: How long from submission to decision?

The Submission Process

Most journals use online submission systems (ScholarOne, Editorial Manager, etc.). You’ll upload your manuscript, figures, supplementary materials, and a cover letter explaining why your paper fits the journal. You’ll suggest (and sometimes exclude) potential reviewers. You’ll confirm compliance with ethical standards.

Then you wait. Typically 2-8 weeks for the editor to decide whether to send the paper for peer review, then another 4-12 weeks for reviews to come back. The editor’s decision options:

Accept: Rare on first submission (less than 5% at top journals)
Minor revision: Small changes required; the paper is essentially accepted
Major revision: Significant changes or additional experiments needed; re-review likely
Reject: Often with an invitation to resubmit a substantially revised version, or sometimes a final rejection

Most papers that are eventually published go through 1-3 rounds of revision. This is normal, not a sign of failure. Engaging constructively with reviewer criticism almost always improves the paper.

Open Access and Preprints

The traditional publishing model—journals charge readers for access—is shifting toward open access, where papers are freely available and authors (or their institutions/funders) pay publication fees. Many funding agencies now mandate open access publication.

Preprint servers (arXiv for physics/math, bioRxiv for biology, medRxiv for medical research) allow researchers to share findings immediately, before peer review. Preprints accelerate science—the COVID-19 pandemic demonstrated their value dramatically—but they bypass quality control, which means readers must evaluate preprints more critically than peer-reviewed publications.

Common Mistakes and How to Avoid Them

Overclaiming

“Our results prove that X causes Y.” Unless you’ve run a flawless randomized controlled trial with thousands of subjects and unambiguous results, this is overclaiming. Prefer: “Our results suggest,” “Our data are consistent with,” “These findings provide evidence for.” Hedging isn’t weakness; it’s accuracy.

Circular Reasoning in Discussions

“The treatment worked because it was effective.” This says nothing. The Discussion should explain why or how your results occurred, referencing mechanisms, prior work, or theoretical frameworks.

Burying the Lead

Your most important finding should be prominent—in the title, in the abstract, in the first paragraph of the Discussion. Don’t make readers hunt for it. Some writers save their key insight for the final paragraph. By then, many readers have stopped reading.

Ignoring the Audience

Writing for a specialized biochemistry journal is different from writing for a broad-audience journal like Nature or Science. Know your audience’s background and adjust your level of explanation accordingly. Don’t explain PCR to a molecular biology audience, but do explain it for a general science audience.

Writing for Non-Scientists

Scientific writing increasingly extends beyond journal articles. Grant proposals, policy briefs, press releases, public lectures, and popular science articles require translating technical findings for non-specialist audiences.

The principles shift: jargon must be replaced with plain language, context must be provided more extensively, analogies become essential, and the “so what?” question must be addressed directly. The challenge is communicating accurately without oversimplifying—a difficult balance that many scientists never master.

Science journalism plays a critical intermediary role, and scientists who can communicate their own work clearly to the public have an enormous advantage in securing funding, influencing policy, and advancing their careers.

Tools and Technology

Modern scientific writing involves software at every stage:

Reference managers (Zotero, Mendeley, EndNote) organize citations and automatically format bibliographies. They save enormous time and reduce formatting errors.

LaTeX is the standard typesetting system in mathematics, physics, and computer science. It handles equations, figures, and references with precision that word processors can’t match. The learning curve is steep but the output is beautiful.

Collaborative writing tools (Overleaf for LaTeX, Google Docs, Microsoft 365) enable multi-author writing with real-time editing, commenting, and version tracking. Given that the average scientific paper has 5-6 authors, collaboration tools are essential.

Statistical software (R, Python, SPSS, Stata) generates publication-quality figures and performs the analyses reported in the paper. Reproducible analysis workflows—where code and data are shared alongside the paper—are becoming standard in many fields.

AI writing assistants are entering the picture but remain controversial. They can help with grammar, clarity, and structural suggestions, but generating scientific content with AI raises serious ethical questions about authorship, accuracy, and originality.

The Craft Behind the Science

Here’s something that doesn’t get said enough: writing well is a skill separate from doing good science. Brilliant researchers can produce nearly unreadable papers. Clear writers can make modest findings compelling.

The best scientific papers combine both—strong science, clearly communicated. They’re a pleasure to read not despite being scientific but because good science, clearly presented, is inherently interesting. When you understand exactly what was done, why it was done, and what it means, the elegance of a well-designed experiment or an unexpected finding comes through with full force.

Scientific writing is a skill that improves with practice, feedback, and deliberate study. Read papers in your field with an eye toward craft as well as content. Notice which papers you find easy to read and figure out what the authors did differently. Study the style guides. Accept that your first draft will be terrible and that revision is where good writing actually happens.

The goal is never to show how smart you are. The goal is to transfer understanding from your mind to the reader’s as efficiently and accurately as possible. When you achieve that—when a complex finding becomes clear through your prose—that’s scientific writing doing exactly what it’s supposed to do.

Frequently Asked Questions

How long should a scientific paper be?