Manuscript Word Limits & How to Shorten It

Nearly every journal caps how long a submission can be — but the cap is never universal. It varies by journal, by article type within that journal (a full research article and a brief communication in the same title can differ by a factor of five), and by which section of the manuscript the count actually applies to. Authors who don’t check the specific target-journal instructions before drafting routinely end up cutting a finished manuscript by 20-40% under deadline pressure, which is a worse editing environment than trimming as you go. This guide covers where word limits typically apply, what’s commonly seen across major journals and article types, and concrete techniques for shortening an overlong manuscript without weakening it.

Where word limits usually apply — and where they usually don’t

A single quoted “word limit” almost never covers the whole submission package. Understand the scope before you start counting:

Main text is what the stated limit almost always governs — typically the running prose from the introduction through the discussion/conclusion (methods and results included). This is the number authors need to actively manage.
Abstract is governed by its own, separate and usually much smaller limit — commonly in the 150-350 word range depending on the journal and whether it’s structured or unstructured. See CASRAI’s Abstract definition and the worked Research Paper Abstract example for how that count is applied in practice.
References are almost universally excluded from the main-text word count, though many journals cap the number of references instead (a separate constraint, not a word count).
Tables, figure legends, and titles are typically excluded from the main-text count but are usually capped separately — by number of tables/figures allowed, and sometimes by a per-legend word limit.
Supplementary/supporting information is generally uncounted against the main-text limit, which is exactly why it’s the release valve described below.

Because scope varies this much, the single most reliable step is also the simplest: open the target journal’s current “Instructions for Authors” and confirm, for your specific article type, exactly what counts and what doesn’t — don’t assume a template from a different journal, or a different article type at the same journal, carries over.

Typical word count limits, by journal and article type

The commonly cited main-text limits for original research articles at several widely known journals illustrate the range rather than a fixed rule: NEJM’s original articles are commonly cited around 2,700 words, JAMA around 3,000, The Lancet around 4,500, and Nature-family research articles around 3,000, while Cell Press journals commonly allow considerably more (roughly 6,000-8,000 words) given their extensive supplementary-data model. By contrast, PLOS ONE, eLife, Scientific Reports, and several other journals impose no strict main-text word limit at all for original research. Abstract limits follow a similarly wide range — commonly cited figures include around 125 words for Science, 150 for Cell, 200 for Nature, 250 for NEJM, 300 for The Lancet, and 350 for JAMA.

These figures are useful as a sense of the range, not as a substitute for checking the current instructions of your actual target journal — limits change between journal updates, and the same journal often sets a materially shorter limit for Letters, Brief Communications, Correspondence, or Case Reports than for a full Research Article. A journal’s word-count instructions are also not always followed precisely by authors in practice; a peer-reviewed analysis of five major medical journals found real published articles routinely departed from the journals’ own stated instructed limits in both directions, which is a reason to treat the instructed number as a strong target to design toward, not a number every published article in that journal literally hit.

Why the limit exists

Word limits are an editorial and production constraint, not an arbitrary hurdle: they keep print/PDF page budgets predictable, keep peer reviewers’ workload bounded, and — done well — force the kind of prioritization that produces a more readable paper. Treating the limit as a real design constraint from the first draft, rather than an afterthought to solve at submission, produces a better manuscript than treating the eventual cut as pure subtraction from an already-finished draft.

How to shorten an overlong manuscript

1. Trim background and discussion redundancy first

The introduction and discussion are usually the most compressible sections because they’re prose-heavy and often duplicate content by design (the discussion routinely restates results before interpreting them). Concretely:

Cut background context that establishes general field knowledge a specialist reader in that journal already has — cite the supporting literature instead of re-explaining it.
In the discussion, state each result’s interpretation once. A common source of avoidable length is restating a result in full before interpreting it — reference it briefly (“the observed X suggests…”) instead of repeating the number or finding verbatim; see CASRAI’s guide to writing the results section and writing the discussion section for where each kind of content belongs.
Cut speculative extensions (“this could eventually lead to…”) that aren’t supported by the data in hand — these add length without adding evidentiary weight, and reviewers frequently flag them anyway.
Limit the literature review embedded in the introduction to what’s directly necessary to motivate the study’s specific question — CASRAI’s literature review guide covers the fuller standalone version of this task; a manuscript introduction needs a much tighter subset.

2. Move appropriate detail to supplementary materials

Supplementary/supporting information is usually excluded from the main-text word count, which makes it the standard release valve for content that’s necessary for rigor but not necessary for every reader to follow the argument. Reasonable candidates to move:

Extended methods detail (full reagent/instrument specifications, secondary or sensitivity analyses, extended statistical output) — keep the main text’s methods section reproducible in outline, move exhaustive parameter tables out.
Additional or exploratory figures and tables that support but don’t carry the paper’s central claims.
Full survey instruments, interview guides, or coding schemes for qualitative work.
Extended literature or supplementary discussion points a specialist subset of readers would want but that aren’t load-bearing for the main argument.

Don’t over-use this: moving content that’s actually necessary to evaluate the paper’s core claim (e.g., the primary statistical test, the primary endpoint definition) into supplementary material just to hit a word count is a red flag reviewers and editors are trained to catch, and can read as obscuring rather than condensing.

3. Tighten sentence-level prose

Once structural cuts are made, sentence-level editing typically recovers another 10-20% without losing content. Common, low-risk techniques:

Cut redundant qualifier stacking — “it is important to note that,” “it should be noted that,” “in order to” (→ “to”), “due to the fact that” (→ “because”).
Convert nominalizations back to verbs — “conducted an analysis of” → “analyzed”; “made a decision” → “decided.”
Prefer active voice where it shortens the sentence without changing the actual claim of agency — passive voice is sometimes appropriate in methods sections by convention, but it isn’t automatically more objective, and it’s frequently longer.
Cut throat-clearing openers — sentences that spend their first clause restating what the paper is about rather than making a new point.
Replace a defined-once abbreviation’s repeated spelled-out form only after it’s genuinely established — see CASRAI’s research abbreviations style guide for when and how to define abbreviations without adding confusion.
Let a table or figure carry data a paragraph is currently narrating — a well-labeled table replaces sentences that list numeric comparisons one at a time; see CASRAI’s scientific figure rules for journal submission guide for what a journal expects from that table/figure in return.

4. Track the count as you write, using the journal’s own counting rule

Word processors’ built-in word counts don’t always match how a journal counts (some exclude headings, references, or table/figure text; some count differently for LaTeX submissions). Where a journal’s instructions specify what its count includes, count against that definition specifically, not your editor’s default count, and re-check after any late-stage addition — a single added sentence added after the count was last checked is a common, avoidable way manuscripts arrive over budget at submission.

What not to cut

Word-limit pressure is a common driver of underreporting that later causes real problems in peer review or post-publication — cutting content that undermines reproducibility or transparency to save length is a false economy. Protect, in particular:

The specific statistical tests used and how multiplicity/missing data were handled.
Sufficient methodological detail for a reader in the field to assess (if not fully replicate) the study — move detail to supplementary material rather than deleting it outright.
Ethics, consent, funding, and conflict-of-interest disclosures — these are typically required regardless of the word budget and are usually excluded from the count for exactly that reason; check the journal’s instructions rather than assuming they’re counted against you.
Limitations — a shortened limitations paragraph is a common and visible casualty of late cutting, and reviewers and editors specifically look for it.

If you’re still over the limit

Most journals have a defined process rather than a hard, unappealable wall: some allow a brief cover-letter note requesting an exception with justification (see CASRAI’s guide on the cover letter for journal submission for what that document typically covers), some route an over-length submission to a different article type with a more generous limit, and some simply return the manuscript for revision before it goes to peer review. Contacting the editorial office before submission, if a journal’s guidance is genuinely ambiguous for your case, is normal practice and usually faster than submitting and finding out.

Frequently asked questions

Does the abstract count toward the main-text word limit?

Almost never — journals virtually always set a separate, smaller word limit specifically for the abstract, distinct from the main-text limit. Check both numbers independently; hitting the main-text target doesn’t guarantee the abstract is within its own limit.

Do references count toward the word limit?

Reference lists are almost universally excluded from a journal’s stated word count, though many journals instead cap the number of references allowed for a given article type — that’s a separate constraint worth checking alongside the word count itself.

What happens if a manuscript exceeds the stated word limit?

It depends on the journal: some editorial offices will return the manuscript for shortening before it’s sent to peer review, some allow a justified exception requested in the cover letter, and some route it to a different article type with a higher limit. It’s rarely an automatic, silent rejection, but it is usually a delay — resolving it before submission is faster than resolving it after.

Are journal word limits enforced strictly?

Enforcement varies by journal and by how a manuscript is actually assessed — published research comparing journals’ instructed limits against their own published articles has found real, measurable gaps between the two in practice. Treat the instructed number as the real design target regardless; “some published articles run long” isn’t a reliable strategy for your own submission.

Should figure legends and table titles be included in the word count?

Typically not counted in the main-text total, but frequently capped on their own terms (a per-legend word limit, or a maximum number of tables/figures) — check the specific instructions rather than assuming figure and table text is unconstrained just because it’s excluded from the main count.

For the broader mechanics of manuscript preparation and submission, see CASRAI’s scholarly writing and publishing pillar page, and the companion guides on how to write a research paper and research paper format.