Try Wonders for free
Search for this question and you will find confident answers such as "one source per page" or "at least 50". Almost none cite anything. In the guidance we could find, universities avoid fixed numbers, a few publish rough ranges, and the expectations of your department and supervisor override both. This guide collects the figures that are published, explains why they vary, and gives you a way to judge sufficiency that does not depend on a count.
What universities actually publish
Several library guides say explicitly that there is no figure. The University of California, Santa Barbara library puts it this way: "There is no magic number regarding how many sources you are going to need for your literature review; it all depends on the topic and what type of literature review you are doing." An emerging topic will have few sources, which helps you argue that more research is needed. A heavily studied topic will have many, and you limit how far back you look.
A few institutions do give planning ranges:
| Level | Published guidance | Source |
|---|---|---|
| Undergraduate review | 5–20 titles depending on level | TUS Library (Technological University of the Shannon), literature review guide |
| Honours dissertation | 20+ titles | TUS Library |
| Master's thesis | 40+ titles | TUS Library |
| Dissertation of about 8,000 words | 8–10 significant pieces (books or articles) | University of Kent, Writing a literature review |
| Dissertation of 12,000–15,000 words | Up to 20 major pieces of work | University of Kent |
| Undergraduate final research paper | At least 13 sources beyond the syllabus readings (about one per week of term) | Raul Pacheco-Vega, professor at FLACSO Mexico, describing his own courses ("my totally non-scientific take") |
| Graduate course literature review | 26 sources (two additional papers per week), rising to 39 (three per week) for PhD students | Raul Pacheco-Vega |
Two details matter when reading this table. The Kent figures refer to significant or major works that you discuss in depth, not to the total length of the reference list, which will be longer. And every one of these sources tells students to confirm with their supervisor or lecturer.
For doctoral work we found no published number. The best benchmark is local: open three recent theses from your department and look at their reference lists.
Putting it together
| Context | Planning figure | Basis |
|---|---|---|
| Undergraduate review | 5–20 titles | TUS Library. Pacheco-Vega asks for at least 13 beyond the syllabus in his own undergraduate courses |
| Undergraduate or honours dissertation | 20+ titles | TUS Library. Kent suggests 8–10 significant pieces for 8,000 words and up to 20 for 12,000–15,000 words |
| Master's thesis | 40+ titles | TUS Library |
| Doctoral thesis | No published figure that we could find | Look at recent theses in your department |
| Journal article | Look at recent articles in the target journal | See disciplinary norms below |
| Systematic or scoping review | Not chosen by the author. Determined by the eligibility criteria | PRISMA 2020 |
Use the planning range to schedule your reading. Do not use it as a target to hit or a ceiling to stop at.
The last row deserves a note. In a systematic review you include every study that meets the criteria, however few or many that is. Asking "how many sources" is the wrong question there. The right question is whether the inclusion and exclusion criteria are justified.
Why the number varies
Discipline
Reference lists differ a great deal by field. An analysis of 729 articles from 147 journals across 27 subject areas (dos Santos, Peroni and Mucheroni, 2022) reports that only three subject areas averaged under 30 references per article: health professions (27.10), computer science (28.85) and mathematics (22.76). Dividing the reference totals in their Table 1 by the number of articles gives about 57 per article for the life sciences and about 54 for the social sciences. These are journal articles, not theses, but the point carries over: the same count can be generous in one field and thin in another. The best benchmark is local. Open three recent theses from your department and count.
Type of review
A narrative chapter selects the most relevant and influential work. A scoping review maps the available evidence on a broad question. A systematic review includes whatever meets the criteria. The differences are set out in systematic vs. scoping vs. narrative review.
Maturity of the topic
A topic that emerged recently may have only a few dozen relevant papers in total, and citing all of them is complete coverage. For a mature topic, you cover the reviews and landmark studies and then concentrate on the part of the literature closest to your question.
Word count
Critical engagement takes space. As a rough check, if your review averages more than one new source every 60 to 80 words, you are probably listing. If a 5,000-word chapter rests on twelve sources, it is probably leaning too heavily on a few authors. These are heuristics of ours, not rules from any style manual.
Four tests for "enough"
1. The key-works test
Are the papers that everyone in the area cites in your review? Find them by checking which references recur across the reference lists of your ten most relevant papers, and by sorting database results by citation count. A missing landmark study is an omission that a reader who knows the field will notice.
2. The saturation test
Run a new search with different keywords, or check the reference list of a newly published paper. If nearly everything relevant is already in your library, you have reached what Pacheco-Vega calls conceptual saturation: reading more yields the same themes through different cases. If a new search still turns up unfamiliar authors and ideas, keep going. Keshav's method for literature surveys uses the same stopping logic: keep iterating for as long as the papers you read cite a key paper you have not yet found.
3. The per-theme test
Look at your outline or synthesis matrix and count sources per theme. A theme supported by one or two sources is either a gap worth naming as such or a sign that you have not searched that angle. A theme with thirty sources probably needs splitting.
4. The recency test
Sort your reference list by year. Is the most recent year or two represented? A review that stops a few years before submission usually means it was written early and never updated. Save your searches and rerun them a month before you submit.
Worked example
This is a hypothetical student, and the counts are made up to show the reasoning. They say nothing about the actual literature on the topic. A final-year psychology student has a 3,000-word literature review on sleep and academic performance in university students. The brief gives no number.
- Planning figure. The TUS rule of thumb of 5–20 titles for an undergraduate review, and 20 or more for an honours dissertation, suggests planning for about 20.
- Structure first. Four themes: sleep duration and grades, sleep quality and regularity, mechanisms, and interventions. About 650 words each, plus an introduction and conclusion.
- Allocation. Four or five sources per theme gives 16 to 20. Add two or three review articles that frame the field.
- Searching. A first search produces a few hundred results. Title and abstract screening cuts that to a few dozen, and a first-pass skim (five to ten minutes each) cuts it again to those worth reading properly.
- Tests. The studies that recur in most reference lists are in. A second search with a different keyword adds two papers, and a third adds nothing new. One theme has only two sources, so the student runs one more targeted search before deciding whether to describe it as thinly studied. The newest source is from the current year.
- Result. Somewhere in the low twenties, each source used in the argument. The number is an outcome of the process.
Common mistakes
- Counting instead of covering. Forty sources that miss the two central studies make a weaker review than twenty that include them.
- Padding. Citing five papers for an uncontroversial statement, or adding sources you have read only as abstracts. You should be able to say what each source found and how.
- Citing sources you have not read. Copying a citation from another paper's reference list passes on any error in it. It is also how fabricated references generated by chatbots end up in student work. See how to check whether an AI-generated citation is real.
- Over-reliance on one author or research group. If a quarter of your references share an author, widen the search.
- Only one source type. A review built entirely on websites and textbooks, or one that ignores the books that define a humanities field, signals a shallow search.
- Ignoring the brief. If the rubric says "a minimum of 15 peer-reviewed sources", that is the requirement.
References
- dos Santos, E. A., Peroni, S., & Mucheroni, M. L. (2022). An analysis of citing and referencing habits across all scholarly disciplines: approaches and trends in bibliographic referencing and citing practices. arXiv:2202.08469. https://arxiv.org/abs/2202.08469 — the figures quoted above were read from this open preprint; the peer-reviewed version is Journal of Documentation, 79(7), 196–224 (2023), https://doi.org/10.1108/JD-10-2022-0234
- Keshav, S. (2007). How to read a paper. ACM SIGCOMM Computer Communication Review, 37(3), 83–84. https://doi.org/10.1145/1273445.1273458
- Pacheco-Vega, R. (2017). How many sources are enough? Six questions on breadth and depth of literature reviews. https://www.raulpacheco.org/2017/06/how-many-sources-are-enough-six-questions-on-breadth-and-depth-of-literature-reviews/
- TUS Library, Technological University of the Shannon. Literature review guide: library summary. https://tus.libguides.com/c.php?g=280076&p=1866569
- University of California, Santa Barbara Library. Literature review: strategies to finding sources. https://guides.library.ucsb.edu/litreview/findsources
- University of Kent. Writing a literature review. https://student.kent.ac.uk/studies/literature-reviews/writing-a-literature-review
