Writing a Scientific Manuscript: Structure, Style, and Best Practices

A scientific manuscript is the formal record of original research — the document that carries findings from a laboratory, field site, or data archive into the permanent scientific literature. The conventions governing its structure, language, and ethical obligations are not arbitrary; they exist to make research reproducible, auditable, and comparable across disciplines. This page covers those conventions in detail: what each section of a manuscript must accomplish, where authors commonly go wrong, and how the choices made at the drafting stage affect a paper's fate in peer review and beyond.

Definition and scope
Core mechanics or structure
Causal relationships or drivers
Classification boundaries
Tradeoffs and tensions
Common misconceptions
Checklist or steps (non-advisory)
Reference table or matrix

Definition and scope

A scientific manuscript, in the operational sense, is a structured document submitted for peer review and publication in a scientific journal. It is distinct from a thesis, a preprint, a conference abstract, and a review article — though all of those may share some of its features. The defining characteristic is original contribution: a manuscript reports data, analysis, or theoretical frameworks that advance the existing literature rather than summarizing it.

The scope of "manuscript" covers a surprisingly wide range of document types. Original research articles, short communications, technical notes, and registered reports are all manuscripts. What unites them is the obligation to follow a reproducible reporting standard and to withstand scrutiny from independent reviewers. The peer review process is not incidental to publication — it is the mechanism that elevates a document from a private claim into a recognized contribution to science.

Manuscript preparation standards are set partly by journals themselves, partly by discipline-specific reporting guidelines (CONSORT for randomized clinical trials, PRISMA for systematic reviews, ARRIVE for animal studies), and partly by general norms codified by bodies like the International Committee of Medical Journal Editors (ICMJE) and the Committee on Publication Ethics (COPE).

Core mechanics or structure

The standard architecture for an empirical research manuscript follows what is known as the IMRAD format — Introduction, Methods, Results, and Discussion — a structure that the American National Standards Institute formally adopted in 1979 and that now dominates the natural and social sciences.

Title and Abstract do more structural work than their brevity suggests. The title is indexed by every database that picks up the paper; its keywords determine discoverability. The abstract — typically 150 to 300 words depending on journal requirements — must stand alone as a complete summary because it is often the only part a reader encounters. Structured abstracts, which break content into labeled subsections (Objective, Methods, Results, Conclusion), are standard in biomedical journals and increasingly common in other fields.

Introduction has a specific job: establish the research gap. It is not a literature review. The conventional three-move structure described by linguist John Swales in 1990 — establish a territory, identify a niche (the gap), and occupy the niche — remains a remarkably accurate description of what effective introductions do. The final sentence of the introduction typically states the study's objective or hypothesis explicitly.

Methods is the reproducibility section. Every procedural detail that would be needed for an independent researcher to replicate the study belongs here: sample sizes, materials and reagents (with manufacturer, city, and country), statistical software (with version number), and ethical approvals. The ARRIVE 2.0 guidelines from the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs) specify 21 reporting items for animal research alone — a sense of how granular this section must be.

Results reports findings without interpretation. Tables and figures carry most of the evidentiary weight; the prose guides the reader's attention to what is significant without editorializing. Statistical reporting requires precision: p-values, effect sizes, confidence intervals, and sample sizes should all appear together, not in isolation.

Discussion is where interpretation lives. It connects results to existing literature, acknowledges limitations honestly, and proposes mechanisms without overclaiming. The most common structural failure here is the "mega-discussion" that re-narrates every result before interpreting any of them — a pattern reviewers flag immediately.

References must conform to the journal's citation style. Errors in reference lists are among the most common manuscript deficiencies identified in editorial review, according to COPE's guidance on editorial processes.

Causal relationships or drivers

The structure of a scientific manuscript is not bureaucratic tidiness — it maps directly onto how scientific claims are evaluated. Editors and reviewers use the Methods section to assess whether the study design supports the conclusions drawn in the Discussion. A mismatch between the two is a structural flaw, not a stylistic one, and it is the primary reason manuscripts are rejected outright rather than sent for revision.

The rise of pre-registration — authors publicly documenting their hypotheses and analysis plans before collecting data — has tightened this relationship further. Journals publishing registered reports, including a growing cohort tracked by the Center for Open Science, commit to publication based on the soundness of methods rather than the significance of results, which structurally eliminates outcome-reporting bias from the manuscript's architecture.

Reporting guidelines like CONSORT, PRISMA, and STROBE exist precisely because unguided authors consistently omit information that replication requires. A 2016 analysis published in PLOS ONE found that fewer than 50% of published randomized controlled trials included all 25 items required by the CONSORT checklist — a finding that demonstrates how structural compliance is an active challenge, not a default outcome.

The manuscript submission process itself creates structural pressure: most journal submission systems now require authors to complete checklist forms confirming that reporting standards have been met, which incentivizes completeness at the point of submission.

Classification boundaries

Not all documents submitted to journals are manuscripts in the primary sense. The distinctions matter because they carry different structural expectations:

Original research articles report new empirical data with full IMRAD structure.
Review articles (narrative or systematic) synthesize existing literature; they follow PRISMA if systematic and do not include a primary Methods section in the IMRAD sense.
Short communications / letters present preliminary or focused findings, typically with a 2,000-word ceiling and condensed structure.
Case reports (common in clinical medicine) follow CARE reporting guidelines and center on one or a small number of patient cases.
Methods papers describe a new technique or tool rather than a finding; their "Results" section demonstrates the method's performance.
Registered reports are reviewed in two stages — before and after data collection — making them structurally distinct from any of the above.

The boundary between a short communication and a full article is often defined by the journal rather than the author. Choosing the wrong article type at submission is a common early-stage error that triggers desk rejection before peer review begins.

Tradeoffs and tensions

The IMRAD structure imposes clarity at a cost: it flattens the actual nonlinearity of research. Discoveries made during data analysis that prompted retrospective hypothesis revision, methods that evolved mid-study, and results that surprised the investigative team all get smoothed into a tidy narrative that can misrepresent the scientific process. This tension between transparent reporting and retrospective coherence is discussed explicitly in research integrity literature, including resources published by the Office of Research Integrity (ORI).

A second tension sits between concision and completeness. Journals impose word limits — Nature research articles run to approximately 3,000 words of main text — while reporting guidelines demand specificity that can easily consume that budget. The solution most journals now use is supplementary material: full protocol details, extended datasets, and additional figures go into a supplement, keeping the main text readable. Whether this practice helps or hurts reproducibility is genuinely contested; materials buried in supplements are accessed far less frequently than main-text content.

A third tension involves the relationship between statistical significance and effect size. Manuscripts that report only p-values without effect sizes or confidence intervals misrepresent the practical significance of findings. The American Statistical Association's 2019 statement in The American Statistician called explicitly for moving beyond p < 0.05 as the binary threshold for scientific conclusions — a call that has reshaped how journals in psychology, ecology, and medicine now frame their results reporting requirements.

Common misconceptions

"The abstract is a preview." An abstract is a standalone summary — complete enough that a reader who never accesses the full paper can understand what was done, found, and concluded. Treating it as an appetizer that withholds key findings is a structural error that reduces discoverability and frustrates database indexing.

"Passive voice is always correct in scientific writing." This has been the stylistic orthodoxy for decades, but major journals including PNAS and Nature now actively encourage active voice where it improves clarity. The American Psychological Association's Publication Manual (7th edition, 2020) recommends active voice as the default.

"Longer introductions demonstrate expertise." Introductions that run beyond 600 to 800 words in most empirical articles tend to delay the research question and lose reviewers before they reach the Methods. The purpose of the introduction is a precise gap statement, not a comprehensive literature review.

"The Discussion can claim what the data imply." The Discussion section must stay within the evidentiary bounds of the actual results. Inferring causation from correlational data, generalizing a single-institution sample to a national population, or proposing mechanisms unsupported by the study design are all forms of overclaiming that reviewers are specifically trained to identify. The research ethics and publication standards framework treats this as a form of misrepresentation, not merely a stylistic choice.

"Supplementary materials are as visible as the main text." Studies on reader behavior consistently show that supplementary files are downloaded at dramatically lower rates than the main paper. Methods or data that are critical to interpreting findings should be in the main text or at minimum flagged prominently.

Checklist or steps (non-advisory)

The following sequence reflects the conventional workflow for preparing and submitting a manuscript:

Identify target journal before drafting — journal scope, word limits, and formatting requirements shape every subsequent decision.
Select the appropriate article type (original research, short communication, registered report, etc.).
Confirm applicable reporting guideline (CONSORT, PRISMA, ARRIVE, STROBE, CARE, or equivalent) via the EQUATOR Network.
Draft Methods first — the most detail-dependent section; drafting it early catches missing procedural records while the research is still accessible.
Draft Results with figures and tables constructed before the prose, since data visualizations often reveal the structure of the narrative.
Draft Introduction — the gap statement should be written after Results are clear, so the framing aligns precisely with what was found.
Draft Discussion — identify the 3 to 5 most important findings and interpret each in relation to existing literature.
Write Abstract last, after all sections are finalized.
Complete reporting checklist (e.g., CONSORT checklist) and prepare it as a submission document.
Verify reference list against the original sources — not secondary citations.
Confirm author contributions and competing interests declarations meet journal and ICMJE standards.
Submit via the journal's manuscript management system, attaching the cover letter, checklist, and any required data availability statements.

Reference table or matrix

Section	Primary Function	Common Structural Failure	Relevant Standard
Title	Indexing and discoverability	Missing keywords; overly broad or narrow framing	Journal author guidelines
Abstract	Standalone summary	Withholding key results; treating as preview	ICMJE recommendations
Introduction	Gap identification	Literature review substituted for gap statement	Swales CARS model
Methods	Reproducibility	Insufficient detail; missing statistical parameters	ARRIVE 2.0, CONSORT, STROBE
Results	Findings without interpretation	Mixing interpretation with results; incomplete statistics	APA Publication Manual (7th ed.)
Discussion	Interpretation within evidence	Overclaiming; re-narrating results without interpretation	COPE publication ethics guidelines
References	Attribution and auditability	Format errors; secondary citation of primary sources	Journal style guide (APA, AMA, Vancouver, etc.)
Supplementary	Extended data and protocols	Critical methods buried where readers won't find them	Journal-specific requirements

The full landscape of scientific publishing — from open access publishing to how journals are indexed and how impact metrics are calculated — is covered across the reference materials available through Scientific Journal Authority's main index.