Journal word limits: 252 journals and conferences compared

Writing · 6 min read · By Bibek Howlader, Founder, Labrix

Every venue tells you how long your paper can be, but they don't measure it the same way. Nature Communications gives you 5,000 words and doesn't count Methods. JACC gives you 5,000 words and does count references and figure legends. Antiquity counts everything, affiliations included. If you move a paper from one venue to another and only compare the headline number, you can end up thousands of words over without realising it.

We went through the author guidelines of 252 journals, conferences and grant schemes and recorded each length rule together with what it counts. Each entry was checked against the venue's own guidelines or call for papers in 2026. Where a publisher's site blocks automated reading, we used an archived copy of the official page and noted the snapshot date. This post sums up what we found. Every venue below links to its format page on Labrix.

The short version

• 74 venues set a word limit, 119 set a page limit, and 59 state no length limit at all. Journals mostly count words. Conferences mostly count pages.
• The most common word limit is 3,000 (17 venues, including Science, JAMA and most Nature research journals), followed by 5,000 and 3,500.
• The most common page limit is 8 (30 venues), followed by 12 (18 venues). Nearly all of them are computer science conferences.
• 94 venues state an abstract cap. 150 words is the most common (29 venues), followed by 250 (28). They range from 100 words (American Economic Review, The Journal of Finance) to 400 (Brain).

Limits at popular venues

A blank "Not counted" cell means the guidelines we read don't say what's excluded. Check with the editor before relying on it.

What a limit counts matters as much as the number

Here are the differences we kept running into:

• References. The Nature journals, Cell Press journals and many medical journals leave references out of the count. JACC, Antiquity, Current Anthropology and the Emerald accounting journals put them in. A 5,000-word paper with 60 references can be well over a 5,000-word limit that counts references.
• Methods. The Nature research journals and Cell Press leave Methods (STAR Methods at Cell Press) out of the main-text limit, so you can put the detail there. Nature Communications still caps Methods at about 3,000 words.
• Figure legends. Neuron, Immunity, Molecular Cell and other Cell Press journals count the main figure legends. Cell itself, NEJM and the Nature journals don't.
• References at conferences. NeurIPS, ICML, CVPR, SIGCOMM and most systems venues give you unlimited pages for references. ICRA, IROS, ECML PKDD and AIED count references inside the page limit. ICASSP allows a 5th page, but only for references, funding and ethics statements.

Some venues use other units

• Publication units. Geophysical Research Letters caps Research Letters at 12 "publication units", where one unit is 500 words or one figure or table. Journal of Geophysical Research allows 25.
• Characters. The abstract in Physical Review Letters is limited to 600 characters including spaces, not words.
• Limits on single sections. At Psychological Science, the introduction, discussion, footnotes, acknowledgments and appendices together get 2,000 words, while Method and Results don't count. Journal of Neuroscience limits the introduction to 650 words and the discussion to 1,500.

"No limit" rarely means no limit

59 venues state no length limit. arXiv, the regular articles of Physical Review A, B, D and E, TMLR, JAIR and theory conferences like STOC, FOCS and SODA are among them. Most of them still set expectations:

• JMLR has no hard limit, but papers over 35 pages take longer to review, and papers over 50 pages need a justification in the cover letter.
• TCC has no page limit, but reviewers don't have to read past page 15 in Springer format.
• TMLR says length should be justified by the content, and unusually long papers are reviewed more slowly.

Some limits cost money instead of causing a rejection

Several IEEE and ASME journals allow longer papers but charge for the extra pages:

• IEEE Transactions on Visualization and Computer Graphics: 12 pages, then USD 220 per extra page.
• IEEE Transactions on Automatic Control: 12 pages, plus up to 4 more at USD 125 each.
• IEEE Transactions on Robotics: up to 18 pages at submission, with a USD 175 charge for each page beyond 12.
• IEEE Internet of Things Journal: 8 pages are charge-free. That's the free length, not a hard cap.

Before you submit

1. Read the current call or author guide. Limits change. Conferences in particular update them every year. Our data has a date on it, and so should your check.
2. Count what the venue counts. Take out exactly what the guidelines exclude, and nothing more.
3. Check your article type. The numbers above are for research articles. Letters, reports, reviews and short papers usually have their own, lower limits.
4. Count the abstract separately. Most venues cap it on its own, and a few also limit the main text measured from the abstract onwards.

How Labrix uses this

Every Labrix project has a target venue, and Labrix reads that venue's limits from this same table. The editor shows how close the draft is to that limit, and Fit-to-limit trims the draft to size. If a call for papers changes the numbers, you can enter your venue's current limits for that project, and they take priority over our table.

You can browse all venue formats, or start a paper and pick your venue.

If you spot a limit we got wrong, tell us at support@labrix.app. We'll check it against the guidelines and fix it.