Which Harvard referencing style does your university use?

If two Harvard examples disagree, do not average them and do not choose the one that looks more academic. Find the rule-making source for this piece of work. A module handbook may prescribe Cite Them Right Harvard; another university may publish Harvard Bath, Harvard UWE or its own house version. Those names share an author-date structure but can produce visibly different references.

The University of York's Harvard guide says there is no set Harvard manual and directs students to departmental regulations; it also records a separate variation for Archaeology. The University of Bath calls Harvard Bath its locally adapted interpretation and explicitly warns that Cite Them Right examples must be adapted to Bath style. That is direct institutional evidence that “Harvard” alone is an incomplete requirement.

A complete format specification looks like `Cite Them Right Harvard, Newcastle 2025 sheet` or `Harvard Bath, current library guide` - not merely `Harvard`.

This matters even when the author-date citations look identical. A marker may check quotation locators, how many authors appear before et al., whether an article title has quotation marks, how volume and issue are punctuated, and how an access date is enclosed. Mixing individually plausible rules produces a reference list that matches no guide consistently.

Cite Them Right is a substantial referencing resource with models for many source types and styles. The Open University describes its version as a commonly used Harvard variant. It is a sound choice when your institution explicitly adopts it; it is not a universal authority that overrides a named local guide.

Search from the work outward. The closer a rule is to the assessment or publication, the more likely it controls. Save the actual page or PDF and its date; a remembered instruction such as “use Harvard” is not enough to configure a reference manager or audit a formatter.

When two high-priority documents conflict, record the exact conflict rather than silently choosing. For example: `rubric says Cite Them Right; department guide says Harvard Bath; rubric updated 12 September`. Ask the module owner or editorial contact which governs and keep the answer with the submission instructions. This is one of the few cases where a short clarification prevents a full-document rework.

Check scope as well as name. One programme may use Harvard for coursework but a numbered style for a clinical report. A dissertation handbook may add rules for archival material or require a bibliography beyond the cited reference list. A publisher may accept any consistent style at initial submission but impose house style after acceptance. Do not import a rule from a different deliverable.

• Record the guide title, owner, URL, edition or update date.
• Record whether it governs in-text citations, the reference list, a bibliography, footnotes or all four.
• Record any department exception and the source types it affects.
• Save two official worked examples before opening a generator.

A guide may say only “Harvard” on the cover. You can still identify its operational rules by extracting templates for a journal article, a book and a web page. Do not infer a whole variant from one comma; build a small specification from repeated examples and explanatory notes.

Newcastle University's 2025 Cite Them Right Harvard style sheet is unusually clear. Its in-text examples name up to three authors and use the first author plus et al. for four or more. Its reference list records all authors unless there are more than ten. Journal article titles appear in single quotation marks; webpage templates use `Available at:` followed by the URL and `(Accessed: 26 March 2025)`.

Now compare the local Bath guide. Its web template uses `Available from:` and puts the access date in square brackets, as in `[Accessed 29 June 2020]`. Bath explicitly says its Harvard style differs from the one in Cite Them Right. Both entries remain recognisably Harvard, yet a strict check can distinguish them immediately.

The variant is the rule set demonstrated by the controlling guide. A citation-style file with the same display name is only a candidate implementation until its output matches those rules.

You do not need to compare every character in a hundred-entry bibliography to detect the wrong variant. Start with five rule clusters that vary visibly and recur often. Build a one-page house sheet from the controlling guide.

Copy the guide's punctuation exactly into the sheet, but keep bibliographic facts separate from formatting. The author, title, DOI, publication year and page range come from the source itself. Quotation marks, italics, labels and separators come from the style. If the source year is wrong, perfect Harvard punctuation does not rescue the reference.

Also check that the in-text and reference-list thresholds belong to the same guide. Newcastle's sheet, for example, can abbreviate four authors to et al. in text while still requiring a much longer author list in the reference entry. Applying the in-text shortcut to the bibliography is a rules error, not a harmless space-saving choice.

Title case is another trap because software may import capitalization from a database. A style can ask for sentence case while the stored title arrives in headline case or all capitals. Correct only the capitalization that the guide permits; preserve proper nouns, acronyms and case-sensitive scientific names.

Choose two real sources already used in the document, not clean examples copied from the style guide. The test should force the converter to reveal different rules. Source A should be a multi-author journal article with a DOI, volume, issue and page or article number. Source B should be a web page with an individual or corporate author, a publication/update date, a stable URL and a recorded access date.

This test separates three failures that look similar in a finished bibliography. Wrong facts require metadata correction. Wrong variant requires a different style specification or formatter. Incomplete source requires author research because no converter can infer an absent issue number, date or corporate author reliably. Record which class failed before changing settings.

After both sources pass, convert a small mixed batch of five to ten entries and check sorting, same-author same-year suffixes, duplicates and in-text/reference-list matching. Only then process the complete document. A two-source pass shows the central templates are configured; the mixed batch catches list-level behavior and unusual source types.

A generator preview is not a source of truth. The source proves the bibliographic facts; the controlling guide proves the format. You need both.

Pause automatic conversion when the authority is unresolved, the library guide is inaccessible, metadata is incomplete, or the document mixes more than one intentional style. Also pause if citations are live Zotero, EndNote or Mendeley fields and the proposed workflow would flatten them to text. Formatting can be repeated; rebuilding field links and correcting invented metadata is much harder.

• Do not convert evidence problems. A citation attached to the wrong claim needs source review, not punctuation changes.
• Do not normalize quotations. Verify quoted words and locators before changing their citation wrapper.
• Do not guess missing dates or authors. Use the guide's documented no-date/no-author rule only after checking the source itself.
• Do not mix tool defaults. Configure collaborators and reference managers to the same named variant.
• Do not discard a working copy. Preserve the original DOCX and compare citation-only changes before acceptance.

For a safe conversion record four things in the job note: controlling authority, exact variant/date, exceptions, and two-source test results. Then freeze body prose, tables and substantive content; change only in-text citation formatting and reference-list formatting. Recheck the live fields, list order, orphan citations and the five high-risk clusters afterward.

The citation-correction button below is for that formatting task. It opens an upload flow where you choose Harvard and submit the document for citation and reference standardization. It does not humanize the prose or decide which local Harvard rules govern; identify the variant first and include its requirements in your final acceptance check.

The goal is not a bibliography that looks Harvard-like. It is a document whose citations are factually traceable and consistently match the one Harvard variant that governs this submission.