Preprints and Unpublished Work: What Belongs in Your Reference List.
Waraq · August 5, 2026 · RULE OF THE WEEK
Mostly no, with one moving exception. Most journals keep unpublished work, personal communications, and papers still under review out of the reference list; a manuscript you have submitted but that no editor has accepted belongs in the running text, not the bibliography. The exception shifts by publisher: a preprint carrying a DOI is citable in many reference lists precisely because the DOI gives it a fixed, retrievable address. PLOS ONE states the strict version of the rule directly: "Do not cite unavailable and unpublished work, personal communications, or rely upon retracted research in the reference list."
What counts as unpublished in a reference list?
Anything a reader cannot retrieve and check from the citation alone. That is the whole test, and it is more useful than the word "unpublished" itself.
A reference list is a set of promises: each entry says a stranger with library access could find this exact source and read what you read. An email from a colleague fails that promise. So does a dataset you were shown in a meeting, a manuscript sitting in a coauthor's drafts folder, and a paper you submitted last month that no one outside the review process can open. None of them have a public, stable home, so none of them can be an entry. The label on the source matters less than one question: given only what your citation prints, could a reader get to the source and back.
The status ladder: personal communication, submitted, in press, published
A source climbs a ladder from private to public, and in a strict journal only the top rungs earn a place in the reference list. The lower rungs get named in the text or nowhere.
Start at the bottom. A personal communication, an email, a conversation, an unrecorded talk, never enters the reference list under either major style system. ICMJE and APA both keep it in the running text with the communicator's initials and surname and the most exact date you can give. ICMJE also asks for written permission before you attribute anything to that person.
One rung up sits the submitted manuscript, and here the two systems split. ICMJE treats a paper under review as an unpublished observation: text only, again with written permission, no entry. APA is looser and does allow a submitted paper in the reference list, tagged "Manuscript submitted for publication" in place of a journal name. If your target journal runs on Vancouver, as PLOS ONE and most clinical titles do, follow the stricter line and keep the submitted paper out. The choice of system decides more than punctuation, which is the whole point of knowing whether your journal cites by number or by author and year.
The next rung is in press. Once a journal formally accepts a paper, it becomes citable: the entry carries the authors, year, title, and journal, with "in press" or "forthcoming" standing in for the volume and pages that do not exist yet. Acceptance is the trigger, not submission. The top rung, a published paper with its final coordinates, is the ordinary case your reference manager already handles, down to the six-author cutoff before et al. that Vancouver enforces.
Preprints with a DOI: the gray zone journals handle differently
A preprint is a full manuscript its authors posted to a public server such as arXiv, bioRxiv, or SSRN before, or instead of, peer review. Because a server assigns it a DOI, it clears the retrievability test that sinks a submitted manuscript, and many journals accept it in the reference list on those grounds.
That is where the gray zone opens. The strict clause PLOS ONE prints bars work that is "unavailable and unpublished," and a preprint with a DOI is neither: it is available to anyone and posted for the record. So the blanket "no unpublished work" rule does not automatically catch it. APA gives preprints a normal entry, using the server's name where a journal name would sit and printing the DOI. Other journals want the entry flagged as a preprint, or ask you to keep preprints in the text and reserve the list for peer-reviewed sources. There is no single answer, so open your target journal's author guidelines and read what they say about preprints before you format the entry.
Moving a disallowed citation into the text without losing the credit
When a source cannot go in the reference list, you still credit it, in the running prose, by naming the person, the nature of the source, and the date. The credit survives; only its location changes.
For a private source, write the attribution into the sentence: "in an unpublished analysis (J. Haddad, personal communication, March 2026)." The reader learns exactly where the claim came from and that it is not independently retrievable, which is the honest signal to send. A submitted paper under Vancouver takes the same shape as an unpublished observation with the authors and year in the text. Two practical notes. Get the written permission ICMJE asks for before you attach anyone's name to an unpublished claim. And lean on these attributions lightly: a discussion propped up by private communications and papers no reviewer can read reads as thin, whatever the formatting.
How to audit your reference list for unpublished entries before submission
Read every entry as a stranger would and ask whether you could retrieve that source from what the citation prints. Move anything that fails into the text, or cut it.
Do it as a deliberate pass, not a glance. Search the reference list for the words that give unpublished work away: "submitted," "in preparation," "unpublished," "personal communication," "under review." Each hit is a candidate to relocate or remove. Check that every "in press" entry is genuinely accepted and not merely submitted, since that mislabel is easy to make and easy for an editor to catch. A reference list carrying sources no one can open is exactly the kind of avoidable flaw behind the desk reject you can see coming, and matching a list against a journal's own rule sheet is the sort of mechanical pass that a manuscript review on Waraq runs before a human ever gets tired at midnight. The DOI test cuts through most of the doubt: if the source has one, it usually stays; if it has no retrievable address, it usually goes to the text or the cutting-room floor.