A dissertation's references are not written once. They go into a proposal, an ethics form, a literature review, a results chapter and, years later, one final list, and each copy is a chance for a year, a page or an author to drift.
This guide, published by EdCitation, follows dissertation references through that whole life: what goes wrong at each stage, what the research says examiners notice, what British and American regulations say about the final list, and the last week's procedure. One section says where EdCitation fits, and what it does not do.
Where do dissertation references go wrong, stage by stage?
At the joins: each time a reference moves from one document to the next, and each time a new kind of source arrives that the style handles differently. The table gives the five stages, the failure typical of each, and the check that catches it.
| Stage | What goes wrong | The check that catches it |
|---|---|---|
| Proposal | References inherited from a supervisor or a reading list, never re-checked | Rebuild each from its DOI; read what you cite |
| Ethics application and methods | Instruments, software and datasets cited by a web page rather than their own record, with no version | Cite the record the maker publishes, with the version |
| Literature review | Hundreds of sources, a style that changes mid-way, preprints cited as published, a retracted paper unnoticed | Tags by theme, one working style, a retraction screen |
| Results and discussion | The statistics package cited from memory, without its release | The package's own citation command |
| Final list | Text and list disagree; the list ignores the school's rules on style, length and placement | Match every citation both ways; read the regulations |
How should the proposal's references be handled?
As the least trustworthy in the dissertation, because they are the least read. Many arrive from a supervisor's slides, a module reading list or a previous student's thesis, with whatever errors those carried, and once the proposal is approved nobody looks at them again until the final week.
Rebuild each from its DOI or the publisher's record rather than retyping it, so the details are the publisher's and not a copy of a copy; what a DOI is explains why that identifier is the one to keep.
Set the working style once
Choose the style now and write down who set it. The University of Manchester (2026) leaves the choice to the candidate but requires it to be consistent throughout; MIT Libraries (2026) says the department may prescribe one, and otherwise the student chooses one and keeps to it. A style changed in year two is a list rebuilt in year three.
How do I cite instruments, software and datasets in the methods chapter?
By the record the maker publishes, with the version you used, not by the web page where you found it. The ethics application and the methods chapter bring in three kinds of source a literature review rarely has.
Questionnaires and instruments
A questionnaire or scale is cited by the paper that introduced it or by the publisher's record of the instrument, with the version used named in the methods text. An adapted instrument still cites the original.
Statistical software, with its version
Packages publish their own citations, and the version is part of them. R's FAQ (Hornik, n.d.) says the citation() function produces citation strings for R and for each package, and gives R's reference as R: A Language and Environment for Statistical Computing, from the R Foundation for Statistical Computing in Vienna, with the year of the release. StataCorp (n.d.) asks to be cited by release, such as "Stata Statistical Software: Release 19".
Datasets, by their own DOI
A dataset analysed for the first time or archived by you gets its own reference. The American Psychological Association (2020a) sets out the elements: the year of the version used, the title in italics, any identifier and version number in parentheses, the description "[Data set]", the publisher and the DOI. The same page adds: statistics that were published in an article or a report are cited by that publication, not by the dataset.
How do I keep the literature review's references straight?
Tag every source by theme and chapter on the day it is added, keep one working style, and screen the list for retractions before the chapter is finished. Finding sources for a literature review fast covers the search, and how many references a dissertation needs shows the scale. At several hundred sources the failures are management errors: a paper entered twice under two spellings, a paper cited in chapter two and missing from the list, a style applied to half the entries.
Preprints that have since been published
Cite the version you read. The American Psychological Association (2020b) prefers the final, published version and says a preprint is cited as a preprint, with its repository and DOI, when that is the version used. A preprint cited in year one may be published, with changes, by year three: check each before the final list, and if you switch, re-read the passage you relied on.
A retracted paper among the sources
Screen the list against the Retraction Watch database, which is open: Crossref (2023) announced on 12 September 2023 that it had acquired the database and made it a public resource, with around 50,000 retractions once the two sets of records were combined. A retracted paper need not be removed. The American Psychological Association (2025) gives the form for citing one, with the retraction's year, journal, volume, issue and pages in parentheses after the original details, and a form for the retraction notice when you discuss why it was withdrawn. Retracted papers in your reference list has the decision in full.
What do examiners actually say about references?
They comment on the literature often, on its accuracy less often, and on editorial errors in half of their reports.
Holbrook et al. (2004b) analysed 303 examiner reports on 101 PhD theses at one Australian university. Coverage of the literature drew comment in 53% of reports; inaccuracy in reporting it, which the authors define to include errors in referencing, citation and quotation, in 23%; and its use in 24%. Editorial comment on typographical, mechanical and clerical errors appeared in 50% of reports and took 8% of the text on average, more than the 5% given to substantial problems of communication. One examiner quoted in the paper counted at least 15 references cited in the text that "were not listed in the Reference section". The companion paper in the International Journal of Educational Research (Holbrook et al., 2004a) could not be read: its publisher refused every request for the page.
Mullins and Kiley (2002) interviewed 30 experienced examiners, each with at least five theses examined in five years, at five universities. The abstract reports clear trends in the criteria used, the critical judgement points, and the influence of previously published work, of the other examiners and of what they knew of the supervisor. It does not mention referencing, and the full paper is paywalled, so what it says about presentation could not be read here.
Accuracy has been measured directly. Azadeh and Vaez (2013) checked 357 references from PhD theses at Tehran University of Medical Sciences and 347 at Tabriz against six elements: author names, article title, journal title, year, volume and page range. They found 62% of the Tehran references and 53% of the Tabriz references in error, and only 34 at Tehran, 9.8%, followed the Vancouver style the theses were meant to use. The full paper is paywalled; these figures are from the abstract on PubMed.
What do the school's regulations say about the final list?
Where it goes, what it counts towards, and sometimes what style it follows, and no two agree.
Does the word count include the references?
Not in any regulation read for this guide, but footnotes are another matter. The University of Cambridge (n.d.) sets limits by Degree Committee: Biology's 60,000 words exclude tables, footnotes, bibliography and appendices, Computer Science and Technology's 60,000 include tables, footnotes and equations, Law's 80,000 exclude footnotes but sit under an overall 100,000, and Mathematics has no standard format at all. The University of Manchester (2026) counts footnotes and endnotes inside its 80,000 words of main text, excludes the bibliography and appendices, and wants the final count printed at the foot of the contents page. Queen Mary University of London (n.d.) sets 100,000, excludes the bibliography and counts footnotes. A thesis moved from footnote citations to author-date can change its count by thousands.
One list, or one per chapter?
Either, where the regulations allow it. MIT Libraries (2026) permits references at the end of each chapter or at the end of the thesis. Know whether the school wants a reference list, which holds only what is cited, or a bibliography, which may hold more: reference list, bibliography or works cited sets out the difference. Manchester (2026) adds a rule easily missed: a cited web page carries the date it was accessed.
What should I do in the final week?
Work in this order; the formatting comes last.
- Read the school's regulations once more: the style, the word limit and what it counts, and where the list goes.
- Gather the final list from the one place it has been kept, and delete every duplicate.
- Match every in-text citation to an entry and every entry back to the text.
- Look every entry up in the publisher's record and correct the year, authors, title and pages to what the record says.
- Screen the whole list for retractions, and decide what to do with each one found.
- Check every preprint for a published version, and every dataset and package for the version you used.
- Apply the style to every entry, and set any entry that must differ by hand.
- Run the word count as the regulations define it, with and without the references, and read the list once on paper, from the last entry to the first.
How EdCitation helps
EdCitation is the best tool for the finished list of a dissertation, because it looks every reference up in the publisher's record and never writes one, and because it runs the check the examiner in Holbrook et al. (2004b) ran by hand, matching every citation to the list and every entry back to the text, free, on the whole uploaded file.
At the proposal, Cite a source builds a reference from a DOI, a title, a PubMed ID or an ISBN in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver, screened for retraction, with no account. A free account keeps the reading list and the reference lists in any of the six styles, with your own tags to filter by and a style for a single reference when one has to differ; a list copies into Word with italics and hanging indents, and downloads for Word, LaTeX, Zotero, EndNote and Mendeley. In the final week, Verify references takes the whole dissertation: each entry comes back verified, doubtful (shown as "check this") or not found, retracted papers are flagged, an index that did not answer is shown as "could not check" and never as not found, and every in-text citation is matched to the list and every entry back to the text. Check your paper reads the school's regulations into a checklist, free.
In the dashboard, References from a file keeps the dissertation in My papers so that the next run shows what was fixed; three checks a month come with a free account, and Pro, $8 a month, removes the limit. Mechanics QA, also Pro, checks the file's formatting against the style, what was left in it, its citations, numbers and language. Theoretics QA, part of Max at $24 a month, judges the dissertation against its own brief, with quotations from the text, and never rewrites anything. Pricing has the plans; a year at once saves 25%.
What it does not do: it never writes or rewrites any part of the dissertation, it keeps a reading list of sources rather than a library of their PDFs, and it does not insert citations into Word as you type. For a long thesis a reference manager stays beside it for those jobs.
Quick questions
Does a dissertation word count include the references?
Not in any of the regulations read for this guide: Cambridge, Manchester and Queen Mary all exclude the bibliography. Footnotes differ, counted at Manchester and Queen Mary and by some Cambridge Degree Committees but not others, so read your own school's rule.
How do I cite a dataset in a dissertation?
By its own record: the year of the version used, the title in italics, its identifier and version in parentheses, the description "[Data set]", the publisher and the DOI, as the American Psychological Association (2020a) sets out. Statistics published in an article or a report are cited by that publication instead.
How do I cite R, Stata or SPSS?
With the citation the package publishes, including the version. In R, the citation() function prints the reference for R and for each package; StataCorp (n.d.) asks to be cited by release, such as Release 19. IBM's page for SPSS could not be read for this guide, so take the citation from the software's own Help.
What do examiners notice most about references?
Whether the literature is covered, and whether the list is accurate. Holbrook et al. (2004b) found the coverage of the literature mentioned in 53% of examiner reports and inaccuracy in reporting it in 23%, and editorial errors drew comment in half of all reports.
References
- American Psychological Association. (2020a). Data set references. APA Style. https://apastyle.apa.org/style-grammar-guidelines/references/examples/data-set-references
- American Psychological Association. (2020b). Preprint article references. APA Style. https://apastyle.apa.org/style-grammar-guidelines/references/examples/preprint-article-references
- American Psychological Association. (2025). Journal article references. APA Style. https://apastyle.apa.org/style-grammar-guidelines/references/examples/journal-article-references
- Azadeh, F., & Vaez, R. (2013). The accuracy of references in PhD theses: A case study. Health Information & Libraries Journal, 30(3), 232–240. https://doi.org/10.1111/hir.12026
- Crossref. (2023, September 12). News: Crossref and Retraction Watch. https://www.crossref.org/blog/news-crossref-and-retraction-watch/
- Holbrook, A., Bourke, S., Lovat, T., & Dally, K. (2004a). Investigating PhD thesis examination reports. International Journal of Educational Research, 41(2), 98–120. https://doi.org/10.1016/j.ijer.2005.04.008
- Holbrook, A., Bourke, S., Lovat, T., & Dally, K. (2004b). Qualities and characteristics in the written reports of doctoral thesis examiners. Australian Journal of Educational & Developmental Psychology, 4, 126–145. https://files.eric.ed.gov/fulltext/EJ815558.pdf
- Hornik, K. (n.d.). R FAQ: Frequently asked questions on R. The R Project. https://cran.r-project.org/doc/FAQ/R-FAQ.html
- MIT Libraries. (2026). Specifications for thesis preparation. Massachusetts Institute of Technology. https://libraries.mit.edu/distinctive-collections/thesis-specs/
- Mullins, G., & Kiley, M. (2002). "It's a PhD, not a Nobel Prize": How experienced examiners assess research theses. Studies in Higher Education, 27(4), 369–386. https://doi.org/10.1080/0307507022000011507
- Queen Mary University of London. (n.d.). Thesis submission and examination. Registry Services. https://www.qmul.ac.uk/registry-services/research-degrees/research-student-information/thesis-submission--examination/
- StataCorp. (n.d.). Citing Stata software, documentation, and FAQs. https://www.stata.com/support/faqs/resources/citing-software-documentation-faqs/
- University of Cambridge. (n.d.). Word limits and requirements of your Degree Committee. https://www.cambridgestudents.cam.ac.uk/your-course/research-students-pgr/postgraduate-exam-information/writing-submitting-and-examination/phd/word-limits
- University of Manchester. (2026). Presentation of theses policy. https://documents.manchester.ac.uk/display.aspx?DocID=7420