How to Cut a Literature Review Down to the Word Limit (2026)

At the University of Manchester’s Global Development Institute, the dissertation count is "12,000 to 15,000 words (plus or minus 10%)" and it includes "chapter footnotes and endnotes" while excluding "references, interview transcripts and abstracts" (GDI dissertation handbook, 2023/24). A student on that programme who trims the bibliography saves nothing, and one who hides a sentence in a footnote saves nothing either.

What does a dissertation word limit actually count?

A word limit is the maximum number of words an institution will accept in the assessed body of the work, and each institution defines that body differently. Footnotes, quotations, tables and appendices are sometimes inside the count and sometimes outside it, which is why two students with identical drafts can be over by different amounts.

Here is what four published sets of rules say, as of September 2026. Every row links to the document the figures come from.

The practical reading of that table is that the same edit has different value at different institutions. Moving 400 words into footnotes clears 400 words at ANU and clears nothing at Cambridge HPS, Manchester or SOAS. Check your own programme handbook before you start cutting, because a faculty rule usually overrides the university-wide one.

How far over the word limit can I go before it costs marks?

Tolerance is institution-specific and it is usually written down, so the answer is in your handbook rather than in convention. Three published positions, as of September 2026: the ANU College of Arts and Social Sciences applies "no penalty ... for the initial 10% excess words" and then a 10% penalty; SOAS deducts 5 percentage points for up to 10% over, rising to a mark of zero beyond 30% over (2018/19 policy document); the Manchester GDI handbook sets hard stops at 10,800 and 16,500 words, outside which a dissertation "will not be marked".

One detail catches people out. UCL’s Institute of Education explains that a single figure such as "2000 words" carries "plus or minus 10% leeway", but a stated range such as 1500 to 2000 words has "no 10% extra leeway" (UCL IOE, understanding assessment requirements). A limit written as a range is already the tolerance.

For a PhD thesis the consequence is procedural rather than a deducted mark. Cambridge requires a statement of length on submission, and warns that unless permission to exceed the limit was obtained beforehand, a thesis over the limit "may not be examined until its length complies with the prescribed limit" (Cambridge word limits).

What should I cut first when a literature review is over the limit?

Cut in this order: paragraphs that describe a study without using it, then repeated descriptions of method that are not the point of your own study, then quotations that could be paraphrased, then the background section that repeats your introduction. What you do not cut first is the number of sources, because at most institutions the reference list sits outside the count anyway.

What each kind of cut is worth depends on the rules in the table above, which is why the order matters.

The first pass that reliably finds words is a claim audit. Go through the chapter and mark every paragraph with the claim it supports in your argument. Any paragraph you cannot label is a paragraph describing a study for its own sake. A chapter written study by study spends a paragraph on each source, while a chapter written claim by claim spends a paragraph on each idea and cites three or four sources inside it, so the same reading arrives in fewer words. That is the arithmetic behind a theme-based structure.

Can I move words into footnotes or an appendix to get under the limit?

Usually not, and the rules say so in writing. Footnotes are counted at Cambridge HPS ("including footnotes but excluding the bibliography"), counted in the Manchester GDI handbook as "chapter footnotes and endnotes", and counted at SOAS, whose policy defines word count as including "quotations, footnotes, titles, abstracts, summaries and tables of contents". The ANU College of Arts and Social Sciences is the exception among those four, excluding "bibliographies, footnotes and endnotes" from the count.

Appendices are excluded more often, but they are fenced. The Manchester GDI handbook permits "no more than five pages of appendices". Cambridge HPS requires Degree Committee permission to place material in appendices outside the word limit. ANU asks students to declare the length "without appendices, bibliography and footnotes on the final page of their assessment item", which means the appendix is visible to the marker as a separate object rather than a hiding place.

There is a further reason to be careful. An examiner reading a literature review chapter with its evidence in an appendix has to hold two documents in mind to follow one argument, and the synthesis is what is being assessed. Shifting words rarely reads as well as removing them.

How do I compress three description paragraphs into one synthesis paragraph?

Write the claim first, then attach the sources to it, rather than writing one paragraph per source and hoping the claim emerges. Three paragraphs that each summarise a study become one sentence that states what the three agree on, one that states where they diverge, and one that says what the divergence means for your research question. That is the move that gets a long review under a limit without losing evidence, because the citations all survive and only the restatement goes.

Litrevu is an AI literature review assistant that turns the papers a researcher has already gathered into a cited first draft, with every citation traceable to the uploaded source. The draft comes out in the claim-first shape rather than the paper-by-paper shape, which is the form you want when you are trying to say more in fewer words.

The following is a format illustration, not a real output. The bracketed slots are where your own papers and findings go, and the citation tokens stand in for your real references.

Paper by paper, the shape that runs long:

(Author, Year) investigated [topic] using [method] and found [finding]. (Author, Year) also investigated [topic], using [method], and reported [finding]. (Author, Year) examined [related topic] with [method] and concluded [finding].

Claim first, the shape a word limit rewards:

[Claim about the state of the evidence] is supported by (Author, Year) and (Author, Year), both using [method]. (Author, Year) reaches a different conclusion on [specific dimension], which matters here because [link to your research question].

The second shape carries the same three citations in one paragraph instead of three, and it tells the examiner what you think rather than what you read. For the longer version of this technique, including how to handle a study that sits awkwardly across two themes, see how to synthesise sources in a literature review.

How much of the whole document should the literature review be?

For a research proposal, a Unisa guide for staff and students puts the literature review at "about 1/3 to 1/2 of the narrative" and caps the master’s proposal itself at no more than 20 pages, with doctoral proposals running "up to a maximum of 35 pages", a figure the guide says includes the references and figures (Unisa College of Agriculture and Environmental Sciences, proposal writing guide, Prof Jana Olivier, 2019). A proposal literature review that has grown past half the document is over-long by its own programme’s guidance, whatever the total says.

For a dissertation or thesis chapter the proportion is set locally and often not at all. Stellenbosch University’s postgraduate layout guidance, revised February 2022, sets no overall thesis word limit and specifies only that the English and Afrikaans abstracts each run "not more than 500 words" (Stellenbosch formatting and layout guidelines). Where the university sets no figure, the faculty or the supervisor does, and that number is the one that binds you.

Two related guides on this blog carry the length question itself: how long a literature review should be by level, and how to write a literature review for a research proposal, where the proportion question bites hardest.

How do I know when I have cut too much?

Three tests, and a chapter that passes all three is short rather than thin. Every claim still carries at least one citation. Every theme still shows at least one study that disagrees, because a review where all the evidence points one way usually means the dissenting work was cut rather than that no dissent exists. Every source in the reference list is still cited in the body, since an uncited reference left behind by heavy editing is a marking problem of its own.

Worth knowing before you panic about being under: a minimum is not always set. Cambridge’s HPS PhD guidance states no minimum length for the thesis, while the Manchester GDI handbook fixes a hard floor of 10,2,000 words and will not mark a dissertation below it. Where no minimum is published, "too short" is a coverage judgement rather than a word count, and the thing being judged is whether a major strand of the literature went missing.

If a cut leaves a theme with one supporting study and no contrast, the honest repair is to restore the contrast and remove a descriptive paragraph somewhere else, not to leave the theme looking unanimous. Losing the disagreement is the expensive kind of cut, and it sits alongside the other faults in 10 common literature review mistakes. Distance students working without a campus writing centre can also check the structure question against the Unisa dissertation literature review guide, and the chapter-by-chapter allocation question against research paper structure.

Common questions about cutting a literature review

Does the reference list count towards my word limit?

At most institutions it does not. The Manchester GDI handbook excludes "references, interview transcripts and abstracts" from the dissertation count, Cambridge HPS excludes the bibliography, and the ANU College of Arts and Social Sciences excludes bibliographies. Check your own handbook, because exclusion is common rather than universal, and cutting sources to save words is usually the wrong first move.

Can I go 10% over my dissertation word limit?

Only if your institution says so in writing. The ANU College of Arts and Social Sciences applies no penalty for the first 10% of excess words and a 10% penalty after that, while SOAS deducts 5 percentage points for up to 10% over. UCL’s Institute of Education adds that a limit written as a range, such as 1500 to 2000 words, carries no extra leeway at all.

Do quotations count towards the word limit?

Usually yes. The SOAS policy defines the count as including "quotations, footnotes, titles, abstracts, summaries and tables of contents", and none of the four documents compared in this guide excludes quotations. Paraphrasing a long block quotation is therefore one of the few edits that reliably reduces a counted total.

What is the fastest way to find 1,000 words to cut in a literature review chapter?

Label every paragraph with the claim it supports, then delete the paragraphs you cannot label. Those are the paragraphs describing a study rather than using it, and they tend to sit in the background section, which most people draft before the argument has settled. A claim audit finds whole paragraphs, which is why it beats a line-by-line tidy-up when you need to lose a thousand words.

Will my marker notice that I moved text into an appendix?

Often, because the rules are built to make it visible. The ANU College of Arts and Social Sciences asks for a declared length excluding appendices, bibliography and footnotes on the final page, the Manchester GDI handbook allows no more than five pages of appendices, and Cambridge HPS requires Degree Committee permission for appendix material outside the word limit.

Is there a minimum length for a literature review?

Some programmes set one and others do not. The Manchester GDI handbook will not mark a dissertation below 10,2,000 words, while Cambridge’s HPS PhD guidance states no minimum length at all. Where no minimum is published, the binding constraint is coverage: a review that omits a major strand of the literature is short in the way that costs marks.