Is it unethical if we cite a paper with mentioning the publication venue (e.g,. author, "paper title," journal name, year) when quoting or mentioning some idea from it, while we have not even seen the paper in its final published form and what we have read is indeed previous versions of it (e.g., because we got it from the author's homepage, arXiv, etc., which are not copy-editted and may have some differences with the final version)?

If yes, what is the solution? (Having many references without any publication venues and e.g., by mentioning the url of the file we have, makes our paper ugly)

  • 1
    All journal style guides I've seen allow for compact citing of papers on arXiv and similar using the identifiers they provide (which are really not much longer than a volume/issue format), you don't need full URLs or anything, so I think that concern of yours makes little sense. I think there is still a good question here about citing preprints when you don't have access to the final version.
    – Bryan Krause
    Jan 22, 2019 at 17:25
  • But I think not mentioning the venue names and just citing many arXiv papers is odd. Have you seen such a paper!?
    – Shayan
    Jan 22, 2019 at 17:57
  • 1
    arXiv is a venue.
    – Bryan Krause
    Jan 22, 2019 at 18:00
  • But, here, by venue I mean e.g., conference or journal name. Have you ever seen a paper that has e.g., 100 references and 80 of them do not have any conference or journal names? I think it's against the norm.
    – Shayan
    Jan 22, 2019 at 18:05
  • 1
    What is the solution? — Read the final published version.
    – JeffE
    Jan 23, 2019 at 11:37

5 Answers 5


I'm going to answer from a (theoretical) computer science perspective.

For historical reasons, in computer science, the "published version" of the paper is likely to be much lower quality than the arxiv preprint. Conference proceedings typically have terrible page limits. Usually 8-12 pages are allowed, but I have heard of 4-page limits. Formally, the published version is not a full paper, it is an "extended abstract".

The standard practice in has become to cite the conference paper but to read the full version on arxiv. It's the only thing that makes sense given our publication system. I rarely bother to find the published version of papers I cite, even if they are readily available.

If it's standard practice in computer science, I struggle to believe that it is unethical in other fields. The arxiv version should be good enough and, if it isn't, it's the authors' responsibility to update it (or withdraw it).


I would be careful of citing one version while referring to a result in another. It has happened before that an ArXiv version contained a result that a preliminary version did not, or that the ArXiv version indicated a problem/bug with the result in the older version.

The reason one should be accurate about citations is because they are essential to understanding the context of your work! If you don't cite accurately, you may be shooting yourself in the foot.

Think of this from the perspective of a future grad student reading your work and trying to retrace your steps. If you refer to a version that does not contain the updated result then the student will be (rightfully) confused by the situation. They may disregard your work, or misinterpret your findings.

I had a similar situation happen to me: we had an updated version on ArXiv with a much better result. A paper came out that rediscovered the same result - I was the one reviewing it, and had to point out to them that the ArXiv version already contained that result...

  • This approach might result in only citing arxiv, if that is the version people find most useful. That's not really a problem, but it makes you wonder what the point of publishing is if no one cites or reads it.
    – Thomas
    Jan 23, 2019 at 6:17
  • But this may result in a paper that has e.g., 90 references, among which 80 are arXiv or homepage papers. Do you think an editor accept that? Or, even do the authors themselves feel happy with that?
    – Shayan
    Jan 23, 2019 at 6:19
  • It’s not about feelings, it’s about proper citation of sources
    – Spark
    Jan 23, 2019 at 7:33
  • If results appear both in ArXiv and conference cite the conference. If just in ArXiv, cite ArXiv!
    – Spark
    Jan 23, 2019 at 7:35

Citing the published work is to be preferred and to do so you should have read them.

In case this is a problem due to paywalls, there are several possibilities:

  1. Ask the authors to for a copy. Most publisher provide a pdf to the authors that they can (and are supposed to) share.
  2. Ask the authors to update their preprint. Some publisher allow this and sometimes it is even requested.
  • 1
    Do really all the authors who cite some papers in their papers have access to the final version of each paper (possibly pay for them), and they see themselves responsible for reading the latest version of each paper, before they cite it? I guess many of them, if not most of them, don't. According to this can we say that it's acceptable and ethical to do so because many, or most people, do that?
    – Shayan
    Jan 22, 2019 at 17:56
  • If someone asked me for a copy of my paper that is available on arxiv, I'd either ignore the email or send them a LMGTFY.
    – Thomas
    Jan 22, 2019 at 18:02
  • @thomas That's kind of rude (unless the arxiv page states that the arxiv version is identical to the published one and in that case you could just say that in your email). I guess in most cases the arxiv version is not identical to the published one.
    – Dirk
    Jan 22, 2019 at 18:28
  • @Shayan I'd say they should Well, I for one try to read the exact version I am going to cite - it is just too often that the preprint and the published version differ (even the numbering may differ, which make citing the wrong version a headache). As I wrote: people should always cite the version they read.
    – Dirk
    Jan 22, 2019 at 18:31
  • And, what should they do when they do not have access to the final version? Not cite it at all!? (Probably in many cases an email to the authors is ignored)
    – Shayan
    Jan 22, 2019 at 18:37

I don't believe there would be any ethical issue. But it is far better for your work and the work of others who read your work if you cite some final, official, version of it. The other versions may well disappear.

If you are in the stage of paper preparation, you can depend on a period of time to update any tentative references if better versions of the papers become available by the time you need to produce a final version. Your tentative citations are likely good enough for your reviewers, but it is better if you update them for final publication if at all possible.

But if an author publishes something, even informally, you can certainly cite it. If you cite things found on the web, however, and for some other places, include the date at which you retrieved the information in the citation. This lets future readers know that things may have changed.

But you should only cite something to which you actually have access and have read. If you read A but cite B there may be a minor ethical issue, but a larger practical issue. Things may have changed and you don't know about the changes. You could wind up both looking foolish and misleading people.

  • 1
    Can you clarify whether you are suggesting they cite the preprint or the final version? My interpretation of your answer is that you are saying it's okay to cite the preprint when a more recent/final published version exists; I think the OP is asking if it's okay to cite the final published version when they have only read the preprint - those are different questions.
    – Bryan Krause
    Jan 22, 2019 at 17:31
  • The question was about the ethics @BryanKrause, and I don't see a problem. But I also answered about the advisability, which I think is problematic. I'll update a bit to clarify.
    – Buffy
    Jan 22, 2019 at 17:40
  • At least in the case of arXiv, even if the version you cite isn't the "final official" version, you still have pretty strong assurance that it won't disappear. Jan 23, 2019 at 1:51
  • "But you should only cite something to which you actually have access and have read." This goes against standard practice in mathematics, where people do cite a lot of sources they haven't ever managed to locate. The honest solution is to be explicit about it ("We were informed that the theorem was also proven in [Antedeluvius 1839], but have not seen this source" or such). Likewise, you can cite a journal paper but refer to "[15, arXiv version, Theorem 3.2]" from the main text of your article. If editors or referees mangle such references, it is their fault ... Jan 23, 2019 at 1:53
  • ... (though, to my knowledge, this doesn't happen near as often as editors mangling bibliographies). Jan 23, 2019 at 1:54

Yeah. Just cite the authors and as much detail about the paper as possible. If you know the eventual journal, say "submitted" or "in press" or whatever describes the status. (These words make it obvious that you did not have the final version. Also it covers you if the thing never makes it through.)

  • I've never seen a reference cited as submitted or in review anyplace but a CV for someone's own papers.
    – Bryan Krause
    Jan 22, 2019 at 20:03
  • 1
    @Bryan Krause: I've seen "submitted to" many times, maybe 5% to 10% of all published papers I've seen (in mathematics). Nearly always these will be to a publication by the author (or one of the authors, as the case may be), but I have seen a very occasional reference to someone else's paper as submitted (more often, in the case of another author's paper, it will be something like "manuscript" or "preprint"). Also, I did a google scholar search with a well known math notion and got quite a few hits. Jan 22, 2019 at 23:25
  • @DaveLRenfro Thanks for the correction; seems like it might be fairly specific to math, though.
    – Bryan Krause
    Jan 22, 2019 at 23:34
  • Normal in material science and chemistry in my experience. Even "unpublished" (e.g. for a manuscript from students a couple years ago, never sent in).
    – guest
    Jan 23, 2019 at 1:44
  • 2
    This does not really work. Editors will ask to replace, or have their assistants replace, the arXiv references by journal references for all papers that have been published. If the journals versions differ from the preprints significantly, your citations may easily lose sense. Of course, the editors then get the blame instead of you, but it's not much of a difference for readers :P Jan 23, 2019 at 1:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .