Do I use the journal page number or the PDF page number in a citation?

The two numbers diverge for a simple reason. A journal article that begins on page 447 of an issue is often distributed as a PDF whose first page is page one. Some publishers add a cover sheet, which shifts everything by one more. Copy the number from your viewer and you produce a citation nobody else can follow.

The rule to apply is that a citation should work for a reader who has the printed journal, the publisher's PDF, a repository copy, and a photocopy. Only the printed page number satisfies all four. It is on the page itself, usually in the header or footer, and it is the number that appears in the article's own bibliographic record.

The reason this matters is the job a citation is supposed to do. Penders states that a citation should identify where an idea originated and let a reader consult and evaluate the cited work directly (Penders, 2018, p. 2). A file position fails that test, because it points at your copy rather than at the publication. Note that none of the published guidance frames this as a numbered rule about offsets, so treat it as an application of the accuracy principle rather than as a citation law you can quote.

Preprints and accepted manuscripts are where this gets genuinely awkward, since they have no printed pagination and their content may differ from the version of record. If you read the preprint, cite the preprint as a preprint, with its own identifier, rather than pretending you read the published article. Authors themselves treat the two versions as distinct. Bourne and colleagues label the pre-peer-review version of their own article as an earlier version rather than as an interchangeable copy of the published one (Bourne et al., 2017, p. 5). If the file you read has no printed pagination, the file's own page number is the correct locator for that version, and your reference has to say which version it is.

Before you submit, check four things against the journal record rather than against your viewer: title and authors, year and journal, volume and issue or article number, and the locator as it appears in the version you cite. Do not assume a reference manager has done this for you. Jin and colleagues report that reference tools carry inconsistencies in bibliographic detail, and that CiteWell's automated validation is centered on PubMed-indexed literature, so anything outside that index needs manual confirmation (Jin et al., 2026, p. 5).

One habit prevents almost all of this trouble: record the page at the moment you take the note, read off the page itself. Reconstructing page numbers at the end of a thesis is slow, error-prone, and the source of a large share of citation corrections.

Do paraphrases need page numbers?

APA does not require them and encourages them when they help a reader locate the passage in a long work. Chicago and MLA generally expect a locator for any specific claim. In a thesis, include them for anything specific, since an examiner checking a claim should not have to search a forty-page article.

The judgment is about specificity rather than about quoting. A claim drawn from one paragraph needs a locator whether or not you used the author's words. A claim about the paper's overall argument does not.

Where the idea runs across several pages, give the range rather than a single page. A range is honest about where the support lies and it is what a reader needs to check.

How do I cite a source that has no page numbers?

Use whatever stable division the source provides: a paragraph number, a section heading, an article number, a timestamp, or a chapter for an ebook. Do not use an ebook reader's page numbers, since they change with font size and device.

Online-first and article-number publishing has made this common. Many journals now assign an article number instead of page ranges, and the correct citation uses that number, with a paragraph or section locator for a specific claim.

Whatever locator you choose, be consistent through the thesis. Mixing paragraph numbers and section headings for similar sources looks like carelessness even when both are individually correct.