How many sources does a research paper need?
There is no correct number of sources, and every page that gives you one is guessing. The real rule is one source per claim that needs support, which means the count falls out of your argument rather than being decided before you write it. A marker checks whether your claims are supported, not whether your bibliography hit a target, so a paper with twelve well-used sources beats one with thirty that were never opened. If your assignment brief states a minimum, that overrides everything below, because a brief is a rule and the rest of this is convention.
You are probably here because you have eight sources and a nagging feeling that eight is not enough. The honest response is that nobody can tell you, and the reason is more interesting than the question. Search this and you get a wall of confident numbers, none of which agrees with the next, and almost none of which is attached to anything you could check.
Nobody agrees, and nobody shows their working
Start with the pages that ought to be authoritative. University writing centres and library guides answer this question constantly, and they contradict each other on the same object. Coventry University's academic writing checklist asks for "at least 6-10 per 1000 words." Baker College Library's answer to the same question, for a four to six page essay, is that "Usually 3-4 reliable sources should be sufficient." Those two are roughly three times apart, and a paper that satisfies one fails the other outright.
Some pages manage to disagree with themselves. Baker's answer opens by stating that "There is no official rule, as every paper is different with varying requirements specified by the Instructor," then supplies a rule one sentence later. Dakota Wesleyan University's citation guide offers two rules of thumb on the same page, one counted per page and one counted per paragraph, and since a page holds several paragraphs the two cannot both hold.
The direction of the advice is not even settled. Some guides treat the number as a floor to clear. Baker warns the opposite way, that "With too many sources, the essay becomes a compilation of opinions from other writers instead of your own." Lancaster University's library, asked how many papers a review should include, answers "The answer is, as many as your search retrieves!" and adds that a search should not be artificially reduced. These are not different emphases on one rule. They are different rules.
What none of them has is evidence. We read through the institutional guides that rank for this question and could not find one that cites a study, a dataset, or any measurement at all for its figure. The two that cite anything cite other guides. Everyone is quoting everyone, and at the bottom of the stack there is nothing.
The commercial pages are worse, and one of them is instructive. An essay-writing service ranking for this query supports its claim about citations per page by attributing it to Glanzel and Schubert (2003) in the Journal of Informetrics. The page carries no reference list and no link. The Journal of Informetrics (ISSN 1751-1577, published by Elsevier) has no registered DOIs before 2007, so there was no 2003 volume for that article to appear in. A page advising students on how many citations to use is supported by a citation that cannot exist. If you want the routine for catching this yourself, we set it out in how to check if a citation is real.
What university guides actually say
Here are the institutional answers side by side, quoted from their own pages. Read down the last column first.
| Institution | What it says | Evidence cited |
|---|---|---|
| Coventry University | "The number of references is at least 6-10 per 1000 words." | None |
| Baker College Library | "There is no official rule, as every paper is different." Then: "Usually 3-4 reliable sources should be sufficient." | None |
| Dakota Wesleyan University | "you will need 1 source for every page." Also, same page: "have about two sources per paragraph." | None |
| University of Southampton | "referencing 30 to 40 sources in your literature review tends to work well," for a dissertation of around 10,000 words. | None |
| University of Kent | "8-10 significant pieces (books and/or articles) for an 8,000 word dissertation, up to 20 major pieces of work for 12-15,000 words." | None |
| Lancaster University | "The answer is, as many as your search retrieves!" | None |
| University of Sussex | "There is no magic number of references. You must give as many references as you need." | None |
| UNC Chapel Hill | "The more controversial an assertion is, the more evidence you may need to provide in order to persuade your reader." | None |
Southampton and Kent are the sharpest pair, because they are answering the same question about the same kind of document. Southampton's suggested range for a 10,000 word dissertation begins above the ceiling Kent gives for a longer one. Both are respectable UK universities, both hedge appropriately, and both are guessing.
The clearest answer in the whole set is also the oldest. The University of Sussex page quoted above was last modified on 16 May 1997 and is still live on a faculty URL. Its full line reads: "An essay is not judged by the length of the bibliography (although an essay without a bibliography either displays dishonesty or a lack of reading)." Twenty-nine years on, that is still better than almost everything ranking above it.
One source per claim that needs support
If the number is not a rule, what is? The workable version is the one UNC Chapel Hill's writing centre states: the more contestable an assertion, the more evidence it needs. That turns a counting problem into a reading problem, and reading your own draft is something you can actually do.
Go through your draft and mark every sentence a reader could reasonably challenge. Not every sentence, and not the connective tissue. Take a paragraph on remote work. "Many organisations expanded remote work after 2020" is background almost nobody disputes. "Remote work reduced commuting emissions in metropolitan areas" is a factual claim with a number behind it somewhere, so it needs a source. "The emissions saving was partly offset by increased household energy use" is more contestable still, so it needs a stronger one, and probably a specific study rather than a summary.
Do that across the paper and the count appears by itself. It is the number of contestable claims you made, plus the sources you cite for definitions, for other people's data, and for the positions you argue against. If that lands at eight, eight is right for your paper. If it lands at twenty-five, the paper needed twenty-five.
This also tells you when a source is doing no work. A reference that appears once, in passing, and could be deleted without changing a sentence is padding, and it reads as padding. Markers notice the paper where source seven is quoted twice, argued with, and returned to in the conclusion, and they notice the paper where sources seven through fourteen each appear once and never again. If you find yourself with a claim and no support, the fix is to go looking, and we cover that in how to find a source for a claim you already wrote.
What has actually been measured
You would expect that somewhere behind all this advice there is a measurement. There is one good one, and it is old.
Staša Milojević's 2012 study in PLOS ONE analysed 213,756 articles from 33 journals across astronomy, mathematics, robotics, ecology and economics, covering 1961 to 2010. On field differences it reports that "Ecology and astronomy currently use around 50 references per article on average, twice as many as the other three fields," referring to the 2006 to 2010 window. That single sentence undercuts every per-word rule in the table above, because it says the dominant variable is your discipline, not your word count.
On the trend, the same paper reports that "the number of references per article had been rising in all five fields," and that "The rate of increase in all fields was similar, with a doubling time of around 30 years."
Reference lists roughly double every thirty years. Milojević measured the doubling time at about that across five very different fields, which means the bibliographies you are being compared against are roughly twice the length of the ones your lecturers wrote as students. A rule of thumb picked up from a writing-help page carries no date on it, and may be calibrating you to a much older norm.
The limits of that study matter as much as its findings. Its data ends in 2010, it covers 33 journals in five fields, and none of those five is the field most students are writing in. Given the growth rate it measured, a figure from that window is now well out of date. Nobody, as far as we can find, has published a current per-field measurement of reference-list length that a student could look themselves up in.
That is a gap we intend to close rather than paper over. The public infrastructure to do it properly exists, and we would rather publish a measurement with its method and its raw data attached than add another confident number to a page that already has too many. Until we have run it, we are not going to tell you what the average is, because we do not know.
A literature review is a different question
The literature review has its own version of this question and it deserves a different answer. In an essay, sources support claims you are making. In a literature review, the sources are the subject, so the count follows from the search rather than from the word limit.
That is why Lancaster's answer, as many as your search retrieves, is more useful than it first sounds. Run a proper search, apply your inclusion and exclusion criteria, and the number that survives is the number. If it is small, that is a finding about the literature and you say so. If it is large, you narrow the question rather than quietly dropping papers to hit a target.
Level-based ladders circulate for this too, stepping up from undergraduate to doctoral work. They are a reasonable intuition, since a thesis does cover more ground than a coursework essay, but the versions published by different universities contradict each other and none of them shows any basis for the steps. Treat them as folklore with a plausible shape. If you are working through a review now, our walkthrough of how to write a literature review covers the search and the synthesis, which is where the real work is.
Why nobody can tell you what student papers cite
There is an obvious better answer to your question, which is not how many references a published article carries but how many a good student paper carries. We went looking for a way to measure that, and hit a wall worth reporting.
Undergraduate research journals exist, they are real peer-reviewed venues that publish student work, and many of them register DOIs. That should make them countable. We checked nine of them, holding 1,408 published works between them, against CrossRef's metadata. The number of those works with a deposited reference list is zero. Small university-press publishers register the article but do not deposit its bibliography, so the public record contains no reference lists for student work at all.
So when a guide tells you an undergraduate paper should have a particular number of sources, understand what that claim rests on. It is not a measurement of student papers, because that measurement does not currently exist in any public dataset. It is somebody's recollection of what they usually see.
When your brief gives a number
All of the above is convention, and convention loses to instruction. If your assignment brief, rubric or module handbook states a minimum, that is a rule and you meet it. This is the one point every guide in the table agrees on, including the ones that refuse to give a number at all: several of them answer the question by telling you to go and read your brief.
If the brief is silent, ask, and ask a specific question rather than a general one. "How many sources should this have" invites a shrug. "I have built the argument on these seven papers, does that look thin for this assignment" gives your marker something concrete to react to, and their answer will be worth more than any figure on the internet, including this page.
Then spend the effort you would have spent counting on the thing that is actually assessed: whether each claim is supported, and whether the support is real. That second part is where papers quietly fail, because a reference that looks right in a list and turns out to be wrong when someone opens it costs more than a short bibliography ever did. CiteOwl is built around that check. It searches real academic databases, reads what it finds, and ties every claim in your draft to the source it actually retrieved, with the quote shown beside the sentence so you can see the source says what the sentence claims. The count then takes care of itself, because every source in the list is there for a reason you can point to.