How Many Sources Should a Thesis Have?

There is no rule that fixes how many sources a thesis needs, and any number you have been told over coffee is somebody's local custom rather than a standard. What does exist is measurement. Researchers periodically count the reference lists of real, accepted theses, and those counts give you something better than a rumour to calibrate against.

A bibliometric study of 153 postgraduate works submitted to the College of Natural Sciences at the University of Dodoma between 2011 and 2023 counted 12,722 citations in total. Split by degree, that comes to roughly 82 cited sources per master's dissertation and 117 per doctoral thesis (Kwigizile et al., 2026, p. 46). Those figures cover one faculty, in one country, in the natural sciences, so treat them as a reference point rather than a quota. They still tell you the order of magnitude: for a master's, dozens rather than a handful, and not many hundreds.

What the counts actually look like

The same study found journal articles dominating the lists, at 46.6% of all citations and 72.9% in the doctoral theses specifically, while monographs, other theses, websites and locally produced material stayed marginal (Kwigizile et al., 2026, pp. 43, 47). Doctoral candidates also leaned harder on current, digital and peer-reviewed material than master's students did.

That composition matters more than the total. A list of 90 items where 70 are blog posts and lecture slides is weaker than a list of 45 built on peer-reviewed articles and two or three standard monographs. Examiners read the shape of a bibliography before they count it.

Your number is an output, not a target

Look at how a documented review arrives at its figure. Thomé and Scavarda searched Web of Science for their review of design-manufacturing integration and got 759 records. Removing proceedings papers, editorial material and meeting abstracts left 499. Restricting to English left 495. Applying their topical exclusion criteria left 42, and all 42 were then read in full text (Thomé & Scavarda, 2015, p. 692).

Nobody decided in advance that the answer was 42. The number fell out of a search string and a set of criteria that any reader could re-run. Farooq Habib's master's dissertation at Cranfield worked the same way and ended with 125 articles drawn from 44 different journals, after a selection process reported step by step (Habib, 2012, p. 53). Habib describes the method's whole point as transparency and replicability, which makes the researcher's decisions and biases explicit instead of hidden (Habib, 2012, pp. 30–31).

So the question to bring to your supervisor is not "how many do I need". It is "is my search wide enough, and are my inclusion criteria defensible". If those two hold up, the count that comes out the other end is the right count for your project.

Before and after

Before. You have been told "about 50 sources" for a master's thesis. You have 31. You spend an evening adding textbook chapters and a few conference papers you have not read, because 50 is the number.

After. You write down the search: databases used, keywords and Boolean structure, years covered, languages, and the reason for each exclusion. You run it, screen the results, and finish with 44 works you have actually read, plus a paragraph in your methodology chapter explaining how you got there.

The second version is six items shorter and considerably stronger. It also survives the question "why isn't X in here", because you can point at the criterion that excluded X.

Bachelor's, master's, doctorate

The gap between degrees in the Dodoma data is smaller than most students expect, 82 against 117 rather than 80 against 400 (Kwigizile et al., 2026, p. 46). A doctorate is not a master's thesis with five times the bibliography; it is a longer argument that has to position itself against more of the existing literature, so the reference list grows with the scope of the claim rather than with the page count.

Field matters at least as much as degree. A theoretical thesis may cite thirty works and be complete. A review sitting across several fields, as Habib's did across general management, operations management, supply chain management, strategy and industrial marketing, has to sample each of them and lands higher (Habib, 2012, p. 11). Ask what a good thesis in your subfield looks like, then open three recent ones from your own department. Most repositories publish them in full, and the reference list is at the back.

Where marks are actually lost

Not in the total. In peer review of journal submissions rather than thesis defences, citation and referencing problems are reported as a contributing factor in 10 to 15% of manuscript rejections, with editors specifically asking authors to remove irrelevant citations, add missing ones, or correct the formatting of the list (Subramanian et al., 2025, p. 100). Your defence delivers that feedback in person instead of by email, but the failure modes are the same ones.

Two of them are worth naming. Padding puts works in the list that the text never uses, which a reviewer spots by checking whether a cited item appears in the argument at all. The other is decay. The same review describes "reference rot", where a cited page either disappears or quietly changes its content between the day you read it and the day someone checks it (Subramanian et al., 2025, p. 100). A bibliography assembled over two years of writing will contain some dead entries by submission day.

That part is worth automating. Every item on your list should resolve to something that exists, with metadata matching what you typed. Paste the list into cytado's source checker and you get back which entries verify against Crossref and the other databases and which do not. If you are still assembling the list, where to find reliable thesis sources covers the databases worth searching first. The four studies cited in this article were themselves found through cytado's corpus, which is why each one carries a real page number.

A working checklist

• Ask your department for the actual requirement. Some programmes publish a minimum; many deliberately do not.
• Use the measured band above as a sanity check, not a goal: dozens for a master's, more for a doctorate, well below both for a bachelor's project.
• Weight the list toward peer-reviewed journal articles and monographs.
• Write your search and selection criteria into the methodology chapter, so the number is defensible.
• Verify that every entry exists and that the year, authors and title match the record.
• Cut anything the text does not actually use.