Choosing a citation style by discipline: a comparison table
Your assignment, department or target journal chooses the style; discipline is only the fallback. This guide maps the common defaults, explains the main citation mechanisms and shows why labels such as Harvard or Vancouver still need a specific local guide.
HumanPen Team
· 6 min read
The short answer, by field
If your department or target journal names a style, that overrides everything below. This is what to assume when nobody has told you.
- Medicine, nursing, public health - Vancouver for most journals following ICMJE recommendations; AMA for JAMA-family and US medical-association titles.
- Chemistry, materials, chemical engineering - ACS.
- Law (UK and Commonwealth) - OSCOLA.
- Engineering, computer science, electronics - IEEE.
- Psychology, education, business, social sciences - APA 7, or Chicago author-date in some social science departments.
- Literature, languages, cultural studies - MLA 9.
- History, art history, philosophy - Chicago notes-bibliography.
- UK, Australia, New Zealand, South Africa, Hong Kong, across many disciplines - Harvard, with the caveat below.
- Chinese universities and journals - GB/T 7714 numbered, or the author-year system where a journal specifies it.
Use the list as a search shortcut, not permission to guess. A computer-science paper may go to an ACM journal rather than IEEE; a medical school may prescribe Harvard for coursework while its researchers publish in Vancouver-style journals. Multidisciplinary journals often publish their own house style that cuts across field conventions.
- First: the exact journal or assignment instructions. Look for a named manual, edition, linked example or journal-specific template.
- Second: department or institutional guidance. This decides local Harvard variants and thesis requirements that a generic web example cannot.
- Third: the disciplinary default. Use the table only when the first two sources are silent, then confirm before final submission.
- Always record the edition or variant. "APA" without "7th edition" and "Harvard" without a named guide are incomplete specifications.
The main citation mechanisms
Styles differ in punctuation, capitalisation and source fields, but the first useful distinction is how a reader connects a citation to its source. There are more than two mechanisms in the styles listed here.
Author-date. The citation names the author and year in the text, and the reference list is usually alphabetical. APA, Harvard, Chicago author-date and GB/T 7714 author-year work this way. Reordering the argument normally does not change reference-list order.
Author-page. MLA normally gives the author and cited page without a year, with an alphabetical Works Cited list. It resembles author-date visually but answers a different question: where in the cited work the passage appears.
Numbered. The citation is a number, usually assigned in order of first appearance, and the reference list follows that order. Vancouver, AMA, IEEE and GB/T 7714 numbered commonly work this way. Moving a citation can renumber every later item, although a live reference manager should perform that renumbering automatically. ACS publications support journal-specific numerical and author-date variants, so the target journal's instructions remain decisive.
Notes and bibliography. Chicago notes-bibliography and OSCOLA put source details in footnotes or endnotes, often with a separate bibliography or tables of authorities. The note number marks a location in the argument; it is not the source's permanent identifier in the way a numbered reference-list citation is.
Do not infer the mechanism from the style name alone. Confirm the target's own example, especially for ACS, Vancouver variants and any house style built on a larger manual.
Harvard is not one style
APA is defined by a manual from the American Psychological Association. MLA is defined by the Modern Language Association. OSCOLA is defined by Oxford’s law faculty. Each has one authoritative version you can check against.
Harvard has none. It is an umbrella term for a family of author-date conventions, and individual university libraries publish their own guides, which differ in punctuation, in whether titles take single quotes, and in whether page numbers carry `pp.`
So a Harvard example you find online can differ from the one your tutor handed out with neither being wrong. Before submitting, your institution’s guide is the authority - which is also why our Harvard page names the variant it applies rather than leaving it implicit.
The differences are not cosmetic when marking is strict. One guide may require the access date for every web source while another uses it only when content is likely to change. One may place the year immediately after the author; another places it later. Corporate authors, missing dates and multiple works from the same year can also be handled differently. A generic "Harvard generator" has to choose a variant even when it does not tell you which.
- Find the guide's owner and revision date. Prefer the library or publisher responsible for your submission over an unattributed summary.
- Match one difficult source. Check a chapter, report or web page rather than a simple journal article; variant differences show up faster.
- Configure the same variant everywhere. Your reference manager, proofreading checklist and collaborators should not use three different meanings of Harvard.
- Save the guide with the project. A stable PDF or link records which rules you followed if the web page later changes.
Vancouver and AMA are not interchangeable
Both commonly number references, use NLM-style journal abbreviations and order the list by appearance. They are still different styles, and three common differences give them away. Individual journal instructions can modify each one, so treat these as comparison points rather than a replacement for the target guide.
- In-text marker. AMA uses a superscript number; Vancouver typically uses a number in parentheses or brackets.
- Page ranges. AMA writes them in full - 245-258. Vancouver abbreviates - 245-58. One character, multiplied across every reference.
- Author cutoff past six. Both list every author up to six. Beyond six, AMA keeps the first three plus et al., Vancouver keeps six.
ACS looks similar again but abbreviates journals with periods - Nat. Methods - following CASSI, where the medical NLM convention uses none. If you submit to both medical and chemistry journals, this is the detail that slips.
Other details are just as visible: DOI presentation, issue-number inclusion, book-edition placement and whether an online-ahead-of-print article carries a date, volume or electronic locator. A converted list can look broadly correct while still mixing rules from two styles. Audit a sample containing several source types instead of checking only the first journal article.
A style name narrows the rules. The journal's own author instructions and examples finish the specification.
Capture sources early; do the final format audit last
Do not wait until the end to collect citation data. Add sources to the reference manager as you research, correct obvious metadata and insert live citations while drafting. That preserves the link between each in-text citation and its library item and lets numbered references update when the argument moves.
What belongs at the end is the final style conversion and quality check: finish structural revisions, settle the target venue, switch to the required style, refresh all fields, then audit the rendered citations and reference list. If citations have already been flattened into plain text, moving paragraphs will no longer renumber them safely.
- Before conversion: resolve duplicate library items, missing authors, wrong item types and incomplete dates.
- After conversion: sample every source type, not every tenth entry; journal articles alone will miss book, chapter, dataset and web rules.
- Before submission: compare the final document with the target's current examples and update Word fields once more.
- If making a one-off output correction: keep the editable source and citation library, because static corrections will not flow back into the manager.
KEEP READING