There is no required number. No UK university publishes one, no national standard exists, and undergraduate marking criteria describe reading in terms of breadth, range and synthesis rather than count. As a working heuristic, a 10,000-word empirical dissertation typically cites somewhere around 40 to 60 sources — but the count is a symptom of adequate reading, never the target.
Why does no one give you a number?
Because the number is not what is being assessed, and any figure would be wrong for most dissertations that used it. A literature-based dissertation in history cites far more than a lab-based project in chemistry that reports original data. A systematic review cites everything that met its inclusion criteria, which might be nine studies or ninety. A dissertation with a narrow, well-defined question cites less than one surveying a broad field — and is often the better piece of work.
The absence is deliberate rather than an oversight. Departments set word limits precisely and reference counts not at all, because a word limit constrains scope while a reference count would reward padding. Any figure circulating online as “the standard number of references for a dissertation” is somebody’s estimate assembled from experience, not a rule you can cite.
What do UK marking criteria actually say?
They talk about reading constantly and never count it. Two named institutions show the pattern.
The London School of Economics Sociology undergraduate mark frame builds its whole ladder out of engagement with literature. First Class Honours (70–100%) requires “an outstanding breadth and depth of reading”, with answers above 80% “outstanding in terms of originality, sophistication and breadth of understanding of relevant themes and material”. An Upper Second (60–69%) “draws on a good range of relevant literature”. A Lower Second (50–59%) “demonstrates some engagement with relevant literature” and “may rely more heavily on description than on independent analysis”. A Third (40–49%) “demonstrates limited knowledge of the relevant literature”. A Fail “may be based entirely on lecture material” and “fail to demonstrate knowledge of the relevant literature”, while a Bad Fail is “likely to be poorly presented with little or no referencing”.
The University of Edinburgh‘s School of Social and Political Science undergraduate marking descriptors draw the same ladder with different words. A-grade work “critically engages with diverse literatures in crafting an original and academically robust argument” and is “grounded in an exceptionally insightful reading of relevant literature”, where “the literature is synthesised in illuminating and creative ways”. Descending the bands, work is “grounded in some engagement with relevant academic literature, although the range may be limited and engagement more in the form of descriptive summaries”, and lower still “the range of reading may, however, be quite limited”.
Read the two frames together and the criterion becomes explicit. The distinction between bands is synthesis versus descriptive summary, and diverse literatures versus a limited range. Neither of those is a number, and neither can be improved by adding entries to a bibliography.
So is there a useful rule of thumb?
Yes, provided you treat it as a diagnostic rather than a target. Sources tend to scale with the length of the argument, and roughly one source per 150 to 250 words of a typical empirical dissertation is a pattern many supervisors would recognise. On the common UK undergraduate word limits that produces:
| Dissertation length | Rough working range | Weighted towards |
|---|---|---|
| 6,000–8,000 words | 25–40 sources | Literature review and discussion |
| 8,000–10,000 words | 35–55 sources | Literature review and discussion |
| 10,000–12,000 words | 45–70 sources | Literature review and discussion |
| Literature-based or review dissertation | Substantially higher | The review is the whole study |
This is a heuristic, not a source. Its only real use is as an alarm: if you are at 10,000 words with twelve references, something is wrong, and if you are at 6,000 words with 180, something else is. Neither number tells you what to fix — the sections below do.
One structural note that changes the arithmetic entirely: a literature-based dissertation redistributes its word budget so that the review and analysis absorb what methodology and results would otherwise take, which is one of the reasons the choice matters at planning stage, as our comparison of a literature review and a systematic review sets out.
Where should the references actually sit?
Not evenly. In a conventional empirical dissertation the density varies sharply by chapter, and markers read the distribution as evidence of whether you understood each chapter’s job.
The literature review carries the most, by a wide margin — it is the chapter whose entire purpose is engagement with other people’s work. The introduction cites lightly, enough to establish that the problem is real. The methodology cites methods literature, not topic literature: the sources justifying your design, your instrument and your analysis. The results chapter often cites almost nothing, because it reports what you found. The discussion is the second densest chapter, because interpreting findings means putting them back into conversation with the literature — and a discussion with no citations is the single clearest sign of a dissertation that stopped at description.
If your reference list looks adequate overall but your discussion cites three sources, the count is hiding the problem rather than revealing it.
What counts as a “good range”?
Four properties, all of which markers can see and none of which is a number.
Recency alongside foundations. A field’s canonical works plus recent empirical studies. All-recent reads as a student who found a database and not a discipline; all-foundational reads as one who stopped reading in 1998.
Primary over secondary. Cite the study, not the textbook’s summary of the study. A reference list dominated by textbooks and review articles tells a marker the primary literature was never opened.
Peer-reviewed at the core. Grey literature, government reports and industry sources have a legitimate place, particularly in applied disciplines, but they supplement the peer-reviewed core rather than replacing it.
Genuine disagreement. Edinburgh’s phrase is “diverse literatures”. A reference list where every source agrees with your conclusion usually means the search stopped when the answer appeared. Including work that complicates your argument, and dealing with it, is what “critical” means in a marking criterion.
What should you do if your list looks thin?
Diagnose before you search. Thin lists usually have one of three causes and each has a different fix.
- The question is too narrow or too new. Widen the search to the adjacent literature — the same phenomenon in a different population, or the same method applied to a different problem — and say explicitly in the review why you are drawing on it.
- The searching was too shallow. One database and one search string produces one slice of a field. Follow the reference lists of your three most relevant papers backwards, and follow their citations forwards.
- You read enough but cited too little. Common and easily fixed: claims stated without attribution, and paragraphs summarising a body of work with a single citation at the end. Attribute properly and the count rises for the right reason.
What not to do is add sources you have not read. Markers open unfamiliar references, a fabricated or misattributed one is found quickly, and a reference list padded with classic works the dissertation never engages with reads worse than a shorter honest one. The same principle governs tool use: our guide to writing a literature review with AI honestly draws the line in the same place — never accept a citation from a tool you have not opened.
And if your list looks too long?
Long is rarely penalised directly, but it carries two real risks. The first is that citations you never discuss make the review a catalogue rather than an argument — exactly the “descriptive summaries” Edinburgh’s middle bands describe. The second is administrative: some departments count in-text citations towards the word limit and some do not, and a heavily cited dissertation can breach a limit for reasons that have nothing to do with its prose, as our survey of dissertation word counts across UK universities found. Check your inclusion rule before you plan chapter lengths.
Keeping the mechanics under control makes all of this cheaper. A reference manager holds the library — our comparison of Zotero, Mendeley and EndNote covers which suits a dissertation — and the formatting rules are in our Harvard referencing guide. If you would rather the bibliography simply maintained itself, Tesify builds it from what your text actually cites as you draft, in your citation style, so the list always matches the argument. It is free to start, and everything stays 100% written by you.
Frequently asked questions
Is there a minimum number of references for a UK dissertation?
No. No UK university publishes a minimum, and marking criteria are written in terms of breadth and engagement rather than count. If your handbook states a figure, that figure governs you — but most do not.
How many references does a 10,000-word dissertation need?
As a working range, roughly 45 to 70 for an empirical project, and materially more for a literature-based one. Treat that as a sanity check rather than a target; a well-argued dissertation at 35 sources outperforms a padded one at 90.
Do markers count references?
Not as a criterion. LSE’s mark frame asks for “an outstanding breadth and depth of reading” at first-class level and “a good range of relevant literature” at 2:1, and Edinburgh’s descriptors separate bands by whether literature is “synthesised in illuminating and creative ways” or handled as “descriptive summaries”. None of those is countable.
Can I have too many references?
Yes, in effect. Sources you list but never discuss turn the review into a catalogue, which is precisely what the middle marking bands describe, and heavy citation can push you over a word limit where your department counts in-text citations.
Should every chapter have the same number of references?
No. The literature review and discussion carry most of them, the methodology cites methods literature, the introduction cites lightly and a results chapter may cite almost nothing. A discussion with no citations is the pattern markers notice most.
Do websites and reports count as references?
They count, and in applied disciplines they can be essential, but they supplement a peer-reviewed core rather than substituting for it. A reference list built mainly from web pages reads as a search that never reached the literature.
Does the reference list count towards my word limit?
It depends on your department, and UK practice genuinely diverges — some count in-text citations, some exclude footnotes and bibliography entirely. Find your own inclusion rule in the handbook before allocating words to chapters.
Is it bad to cite the same author many times?
Not if the author genuinely dominates the relevant literature, which happens in small fields. It becomes a problem when it signals that your reading stopped at one research group — the fix is to find who disagrees with them.
How many sources should be in the literature review specifically?
Most of your total, since it is the chapter whose purpose is engagement with existing work. Judge it by whether you can group your sources into themes with genuine tensions between them, not by how many are in the pile.
My supervisor said my reference list is “thin”. What do they mean?
Usually one of three things: the searching was shallow, the reading was adequate but under-attributed, or the question is too narrow for the literature that exists. Ask which, because the three fixes are entirely different and only one of them involves reading more.
Should I include everything I read, even if I did not cite it?
Only if your department asks for a bibliography rather than a reference list. A reference list contains what you cited; a bibliography may include background reading. Check which your handbook requires, and never pad one to look like the other.
